Quantitative Microscopy: Protein Dynamics and Membrane Organisation

Authors

  • Dylan M. Owen,

    1. Centre for Vascular Research, University of New South Wales, and the Department of Haematology, Prince of Wales Hospital, Sydney, Australia
    Search for more papers by this author
  • David Williamson,

    1. Centre for Vascular Research, University of New South Wales, and the Department of Haematology, Prince of Wales Hospital, Sydney, Australia
    Search for more papers by this author
  • Carles Rentero,

    1. Centre for Vascular Research, University of New South Wales, and the Department of Haematology, Prince of Wales Hospital, Sydney, Australia
    Search for more papers by this author
  • Katharina Gaus

    Corresponding author
    1. Centre for Vascular Research, University of New South Wales, and the Department of Haematology, Prince of Wales Hospital, Sydney, Australia
    Search for more papers by this author

Katharina Gaus, k.gaus@unsw.edu.au

Abstract

The mobility of membrane proteins is a critical determinant of their interaction capabilities and protein functions. The heterogeneity of cell membranes imparts different types of motion onto proteins; immobility, random Brownian motion, anomalous sub-diffusion, ‘hop’ or confined diffusion, or directed flow. Quantifying the motion of proteins therefore enables insights into the lateral organisation of cell membranes, particularly membrane microdomains with high viscosity such as lipid rafts. In this review, we examine the hypotheses and findings of three main techniques for analysing protein dynamics: fluorescence recovery after photobleaching, single particle tracking and fluorescence correlation spectroscopy. These techniques, and the physical models employed in data analysis, have become increasingly sophisticated and provide unprecedented details of the biophysical properties of protein dynamics and membrane domains in cell membranes. Yet despite these advances, there remain significant unknowns in the relationships between cholesterol-dependent lipid microdomains, protein-protein interactions, and the effect of the underlying cytoskeleton. New multi-dimensional microscopy approaches may afford greater temporal and spatial resolution resulting in more accurate quantification of protein and membrane dynamics in live cells.

Roughly one third of the human genome encodes proteins that are membrane-associated either as transmembrane, lipid-anchored, or peripheral membrane proteins. Protein diffusion, described in simple terms by the diffusion constant, D, not only distributes proteins across the membrane but also determines their ability to interact. The forward reaction kinetics of two generic membrane proteins is determined by the sum of their diffusion constants. Hence, the mobility of proteins determines their interaction capabilities in, for example, signalling events. The interaction efficiency is further determined by the local protein concentration which controls the collision probability. In other words, two slowly diffusing proteins interact inefficiently unless their local protein concentration is very high. To construct a complete picture of membrane proteins in a complex cellular process such as receptor signalling, it is not sufficient to know the binding affinities of ligands and kinases—only by quantifying protein dynamics and concentrations can we fully understand interactions and functions.

The kinetic properties of a membrane protein and the types of motion it undergoes are critically influenced by the lateral organisation of the cell membrane. The fluid mosaic model of Singer and Nicolson proposed that the plasma membrane of cells is laterally homogenous—membrane-associated proteins can diffuse freely with Brownian motion in the plane of the lipid bilayer (1) and will therefore be randomly distributed on the cell surface. In this random diffusion model, the diffusion constant, D, is only dependent on the hydrodynamic radius of the protein (which is assumed to be cylindrical) and the viscosity of the membrane and is described by the Saffman-Delbruck model (2). In practical terms, Brownian motion means that the distance a molecule diffuses, measured as the mean square displacement (MSD) in time t in a two-dimensional membrane relates linearly to the diffusion coefficient D:

image(1)

However, cell membranes—and here we focus on the plasma membrane—are complex lipid-protein composites in which protein diffusion becomes anomalous (3) with diffusion coefficients an order of magnitude lower than those observed in artificial membrane systems (4). This retardation is possibly due to protein–protein interactions (5), active transport, cytoskeleton interactions and confinement (6), and membrane lipid microdomains (7). Particularly, lipid rafts, defined as ‘small, heterogeneous, highly dynamic, sterol- and sphingolipid-enriched domains’(8), are hypothesised to facilitate protein–protein interactions by their ability to selectively accumulate proteins and control diffusion via their internal viscosity and/or boundary properties.

In heterogeneous cell membranes, the diffusion coefficient, D, may be divided into local and global values because diffusion over a short distance and time period (for example, within a raft) is clearly different from that over longer distances and timescales (covering many rafts for example). Therefore, by simply changing the observation area and recording time intervals, one can obtain different values for D. Yechiel and Edidin were the first to demonstrate this phenomenon and inferred the existence of protein-rich membrane domains of ∼ 1 μm in size in the plasma membrane of endothelial cells (9). Hence by quantifying the local and global diffusion behaviour of proteins and lipids, it is possible to build more detailed physical models of cell membranes and understand the underlying principles of their organisation. Figure 1 illustrates the link between membrane organisation and protein diffusion in different scenarios that are neither comprehensive nor mutually exclusive.

Figure 1.

Sketch of the primary models describing the microscopic organisation of the cell plasma membrane and their effect on the diffusion behaviour of membrane proteins. Free Brownian diffusion, directed flow, hop diffusion across membrane picket fences, immobile obstacles, and transient confinement in dynamic membrane microdomains.

Accuracy in measuring protein dynamics in live cells has been a major limitation for a long time. Fluorescence microscopy is the only suitable approach to characterise protein dynamics in living cells and in recent years, tremendous progress has been made in hardware, automated acquisition, and analysis algorithms. Today, there are three principle techniques that can be employed for the measurement and analysis of molecular diffusion in cell membranes. These are fluorescence recovery after photobleaching (FRAP), single particle/molecule tracking (SPT), and fluorescence correlation spectroscopy (FCS). The goal of this review is to summarises how FRAP, SPT, and FCS have deepened our understanding of membrane rafts and the global organisation of the plasma membrane.

Brownian Protein Diffusion and Membrane Viscosity

FRAP has been widely used to measure diffusion coefficients in live cells (Figure 2). An area of the sample is first quickly photobleached by intense laser illumination. As mobile fluorophores from outside the bleached region diffuse into the dark area, the fluorescence intensity is seen to recover. The speed of this recovery is related to the fluorophore mobility and thus the diffusion coefficient can be extracted. The fraction of the fluorescence that is recovered after long time-scales is the so-called mobile fraction (Mf) and gives information on the percentage of molecules that are free to diffuse. By using the volume of the laser focal spot for bleaching, one derives more localised diffusion parameters, distinct from the global measurements when the laser spot is used to bleach a larger area such as a disk or rectangle. The size and geometry of the bleached area therefore has to be taken into account when interpreting and comparing FRAP results.

Figure 2.

FRAP. A) a small area of a fluorescent sample is bleached by intense laser illumination. Over time, fluorescence returns as mobile fluorophores diffuse into the bleached volume. B) By plotting the observed fluorescence intensity over time, FRAP experiments are indicative of the diffusion coefficient, D, and mobile fraction, Mf. Post-translational modifications and different experimental conditions such as temperature and cellular cholesterol levels can influence the dynamics of the observed proteins altering D,Mf or both.

FRAP studies were employed to test the hypothesised effects of lipid rafts on diffusion (10). If rafts are small immobile domains of high viscosity, raft-favouring proteins with a stable raft association would have low mobile fractions and diffusion coefficients, and different raft markers would have similar diffusion properties. Proteins with high affinity for membrane rafts including the dual-palmitoylated transmembrane protein linker for activation of T cells (LAT) and farnesylated/palmitoylated H-Ras or proteins that target raft constituents such as cholera toxin subunit B (CTxB) that binds and clusters gangloside GM1 (for a review on raft affinity and targeting see (11)), however, were found to have vastly different diffusion coefficients (see Table 1). It was therefore concluded that the type of membrane anchor, but not raft association per se, determines the diffusion coefficient of proteins in the plasma membrane (10).

Table 1.  Diffusion coefficients and mobile fractions (Mf) for a variety of membrane components and probes measured by FRAP, SPT, and FCS in a range of membrane and cell systems
Bilayer/Cells + TreatmentsProbe-targetDiffusion Coefficient (μm2/s)Mobile Fraction (Mf)Ref.
  1. CHO: Chinese hamster ovary; DMPC: dimysistoylphosphatidylcholine; FITC: Fluorescein isothiocyanate; HUVEC: human umbilical vein endothelial cells; ICAM: intercellular adhesion molecule; LEL-GST: large extracellular loop of glutathione S-transferase; μOR: μ-opioid receptor; NRK: normal rat kidney; PCX: podocalyxin; PH: Pleckstrin homology; RBL: rat basophil leukaemia; SM: sphingomyelin; TMD: transmembrane domain.

FRAP    
DOPC:SM:CHOL 1:1:1 (ordered)FITC-Thy-1 (raft)0.5(12)
Cos-7 cellsKRas-GFP (inner leaflet, non-raft)∼ 1.10.93(10)
 GFP-Fyn (raft)∼ 0.80.93 
 GFP-GPI (raft)∼ 0.60.93 
 Cy3-CTxB (GM1) (raft)∼ 0.1  
MDCKYFP-GPI (raft)0.020.94(13)
 GFP-LAT (raft)0.01/0.09 (RT/37°C)0.93/1.00 
 GFP-PCX-Δtail (non-raft)0.0050.71 
 GFP-EGFR-TMD (non-raft)0.007/0.02 (RT/37°C)0.50/1.00 
HUVECGFP-GPI1.2(14)
 GFP-ICAM-10.42- 
SPT    
C3H 10T1/2Gold bead-Thy-10.080.28(15)
T3T-A14 (37°C)eYFP-HRasC101.13 and 0.290.59 and 0.41(16)
+ mβCDeYFP-HRasC100.95 and 0.140.72 and 0.28 
+ Latrunculin AeYFP-HRasC100.83 and 0.310.67 and 0.33 
T3T-A14 (37°C)eYFP-LckN111.30 and 0.260.84 and 0.16(16)
 eYFP-KRasC141.00 and 0.150.73 and 0.27 
NRK cells (37°C)Gold-μOR0.20 (over 100 ms),(17)
  0.18 (over 200 ms),
  0.12 (over 1.5 s)
 GFP-μOR0.25 (over 100 ms), 
  0.24 (over 200 ms),
RBL-2H3Quantum Dot-Fcε RI0.077(18)
+ Latrunculin A 0.059  
T3T-A14YFP-HRas1.11 and 0.120.75 and 0.25(19)
+ Insulin stimulationYFP-HRas0.96 and 0.100.75 and 0.25 
Cos-7Farnesylated-GFP4.7(20)
Jurkat T cell-anti-CD3 synapse   (5)
Outside CD2 clustersGFP-LAT0.30.36 
Inside CD2 clusters 0.180.64 
Outside CD2 clustersGFP-Lck0.460.41 
Inside CD2 clusters 0.260.59 
Outside CD2 clustersGFP-LckN101.140.53 
Inside CD2 clusters 1.080.47 
FCS    
DOPC (disordered)diI-C186.3(21)
DOPC:SM:CHOL 1:1:1 (ordered)diI-C180.8 
HEK293diI-C181.4(4)
+ mβCDdiI-C180.9 
RBLFITC-CTxB (GM1)Immobile 
+ mβCDFITC-CTxB (GM1)Immobile 
+ Latrunculin AFITC-CTxB (GM1)0.1 
CHO (37°C)GFP-EGFR0.25(22)
HUVECGFP-CD9∼ 0.2(14)
+ CD9-LEL-GST ∼ 0.12  
+CD9-LEL-GSTGFP-CD151∼ 0.15 
  ∼ 0.06  
JurkatGFP-LckN121.7 (raft dependent)(23)
 GFP-Akt0.8 (raft and non-raft 
  dependent)
 GFP-Akt (PH domain)1.2 (raft dependent) 
Cos-7 (confocal FCS)Farnesylated-GFP2.70.59(20)
  7.00.41 
Cos-7 (TIRF FCS)Farnesylated-GFP1.10.72(20)
  5.60.28 

Interestingly, in the apical membranes of Madin-Darby Canine Kidney (MDCK) epithelial cells at room temperature, only the raft-associated proteins LAT and glycosylphosphatidylinositol (GPI)-anchored proteins showed complete fluorescence recovery in bleached regions, whereas, 30% of non-raft proteins were immobile (13). However, the latter became mobile at 37°C. This temperature dependence was interpreted as the formation of a discontinuous liquid-ordered phase at lower temperatures, which segregates and confines membrane proteins. ‘Raft proteins’ can diffuse freely in either phase whereas ‘non-raft proteins’ are excluded from ordered-phase domains which therefore act as obstacles to their free diffusion. Here, FRAP was sensitive enough to detect differences between proteins that could, or could not, dynamically partition into raft domains but only because raft coverage was close to the percolation threshold. It should be noted that there is currently no direct evidence that lipid phase separation does indeed occur in native cell membranes.

Due to spatial resolution restrictions, FRAP studies could not detect the effects of small raft domains on protein diffusion, but they did confirm the general relationship between membrane viscosity and protein dynamics. As expected, diffusion rates for both raft and non-raft markers decreased with decreasing temperature as the membranes become less fluid (10). Goodwin et al (24) observed that cholesterol loading decreased the diffusion rates of Ras proteins due to increased membrane viscosity, but so did cholesterol depletion using methyl-ß-cyclodextrin (mßCD). It is known that the removal of cholesterol, particularly with mßCD not only alters membrane viscosity but can also hinder diffusion (25), possibly due to the induction of gel-like regions (26) and reorganisation of the actin cytoskeleton (27).

In a recent study, the spatial limitation was overcome by simultaneously recording FRAP and anisotropy recovery after photobleaching (ARAP) of GPI-anchored proteins (28). Changes in fluorescence anisotropy are indicative of the fluorescence resonance energy transfer (FRET) between identical fluorophores (homoFRET). While monomers diffused freely, nanoclusters of only a few GPI-anchored proteins were essentially immobile and were non-randomly distributed across the membrane. Together with the concentration-independent and temperature- and cholesterol-dependent nature of conversion between monomers and clusters, the study suggests that the formation of clusters is actively regulated by cortical actin and myosin contractility (28). To which extent protein multimerisation correlates with lipid domains can only be resolved when both processes are recorded simultaneously but independently from each other.

The ability of FRAP to detect the global behaviour of macromolecular complexes was demonstrated for the epidermal growth factor receptor (EGFR) by Lajoie et al (29). By systematically measuring the diffusion rate and the mobile fraction, it was discovered that EGFR association with a galectin lattice overrides the immobilisation of the receptor in caveolae, and hence the tumour suppressor function of caveolin-1. Particularly interesting is the finding that the regulation of EGFR diffusion rates determines its ligand responsiveness. EGFR signalling was suppressed when the receptor was sequestered into immobile caveolae but enhanced, relative to unrestricted diffusion, when the receptor's diffusion rate was retarded by the galectin lattice. The data therefore serve as an indicator that a receptor's mobility, as modulated by the membrane properties, has a strong influence on its interactions and function.

The main limitation of FRAP studies is that sub-populations that diffuse differently cannot easily be identified because FRAP is essentially an ensemble or bulk measurement. In an elegant theoretical paper, Nicolau et alshowed that the existence of high-viscosity islands (rafts) affects the long-range mobility of both raft- and non-raft-partitioning proteins (30). The theoretical model further predicted that all proteins, whether with or without raft affinity, diffuse significantly slower when rafts are present. In this context, it is interesting to note that protein diffusion in cell membranes is several times slower than in artificial bilayer constructs composed of just a few lipid species. For example, the lipid dye diI-C18 has diffusion coefficients of ∼ 1.4 μm2/ s in the cell plasma membranes of human embryonic kidney (HEK) cells (4) but ∼ 6.3 μm2/ s in dioleoylphosphatidylcholine (DOPC)-containing artificial bilayers (21).

In addition to raft domains, diffusion in the plasma membrane of cells may also be hindered by the underlying membrane cytoskeleton and/or the extracellular matrix giving rise to more complex diffusion behaviour with at least two different sub-populations. These systems must be tackled using microscopy techniques with single-molecule sensitivity.

Anomalous Sub-Diffusion: Single Molecule Techniques

To characterise the anomalous sub-diffusion and the relative diffusion rates in different regions of the plasma membrane, SPT is the method of choice (Figure 3). In SPT (31), individual molecules are labelled, successively imaged, tracked, and the trajectories analysed (32,33). When only a few, well separated molecules are fluorescent, their location can be pinpointed with great accuracy by simply calculating the centre of the observed Gaussian point-spread function (34). The localisation in consecutive image frames is then connected to form trajectories and from the mean-square displacement (MSD), the diffusion properties can be derived.

Figure 3.

SPT. Examples of fluorescent particle trajectories for A) free diffusion or directed flow, B) transient confinement in membrane microdomains, and C) confinement and hop diffusion due to the sub-membrane cytoskeleton meshwork. D,E) Effects of different types of motion on plots of the mean-square-displacement (MSD) against time indicating how different types of organisation can be inferred from single-particle trajectories (t = time , v = flow velocity, α, A, B = constants , rTCZ = TCZ size).

The tracking of membrane proteins with latex beads and a laser trap became the first publication that showed that raft proteins are anchored into ∼ 30 nm sized domains that diffuse as an entity across the cell surface (35). Recalling that the diffusion coefficient of membrane constituents is related to their hydrodynamic radius, large tags such as antibody-coated gold beads may significantly alter diffusion characteristics, particularly if protein cross-linking cannot be ruled out. In recent years, most studies have taken great care to ensure that only single proteins are tracked (17) or have exploited the improved signal-to-noise ratio of total internal reflection fluorescence (TIRF) microscopy to directly image fluorescent fusion constructs or small molecule dyes.

The main disadvantages of SPT are that there is an inherent trade-off between spatial and temporal resolution and that statistical analysis is limited unless a large number of molecules are tracked in each experiment. To map the entire cell surface, localisation and connecting molecules into trajectories have to be achieved for a high density of molecules. To date, no standardised tracking algorithm exists that can simultaneously account for high particle density, heterogeneous particle motion, particle interactions (merging and splitting), or temporary disappearance (e.g. blinking). Progress has been made (33,36), but the interpretation of trajectories will still rely on simulations (37) until robust algorithms are shared between researchers.

The advantage of SPT is that it not only allows molecular diffusion coefficients to be calculated but different modes of diffusion of the same molecule can be identified. Monte-Carlo simulations have identified how modes of movement differ from Brownian motion if molecules are actively transported, or confined in immobile or diffusing domains (37).

Recent examples that demonstrate the power of SPT have addressed the influence on protein diffusion of post-translational lipid modifications and raft affinity (5), protein–protein interactions (38,39), and the actin cyto-skeleton (18). Similar to FRAP experiments, SPT was used to assess the contribution of lipid modifications on protein dynamics on the surface of T cells. Palmitoylation of LAT or dual acylation of the Src-family kinase Lck did not alter their diffusion significantly, but ‘trapping’ of these signalling proteins in cholesterol-independent clusters of the co-receptor CD2 did slow them down approximately twofold (5). Such differential modes of movement could only by detected by dual-colour SPT (and not with FRAP). In the case of the T cell plasma membrane, protein–protein interactions have to be factored into the interpretation of modes of motion, which is likely to be the case for other cell types where stimulation of surface receptors results in multi-molecular protein complexes. Dual colour tracking of the GPI-anchored receptor CD59 in the outer leaflet, and signalling proteins anchored to the inner leaflet also revealed how ligand clustering of the receptor induces temporary immobilisations, termed STALLs (stimulation-induced temporary arrest of lateral diffusion) that serve as short-lived platforms for signalling activities (38,39).

Another recent example where SPT revealed the mechanism of receptor dynamics is the tracking of individual Fcε RI receptors with non-bleachable quantum dots (18). Different modes of motion—immobile, free, directed, and confined—were all observed for the same receptor. Confined motion was attributed to the presence of the actin cytoskeleton which was indeed found to dynamically confine the receptor into micron-sized domains.

Transient confinement zones (TCZs), detected by SPT define areas where an observed molecule stays much longer than expected from the average diffusion coefficient and is thought to bear some resemblance to lipid rafts. TCZs are typically 200–300 nm in diameter (40), preferentially trap GPI-anchored proteins and glycosphingolipids (15), and are cholesterol dependent (41). Yet, whether the viscosity differential inside and outside the raft is a sufficient diffusion barrier is questionable (12), particularly since TCZs appear to be temperature independent (41). Lommerse et altracked fluorescent fusion constructs of only the membrane anchor regions of H-Ras (raft), K-Ras (non-raft), and Lck (raft). These three constructs had similar diffusion coefficients (assuming Brownian motion), but two populations of diffusing molecules were observed (16). The major population displayed similar diffusion times to small molecule membrane dyes (0.6–1.6 μm2/ s); however, a second population (16% for Lck and 27% for K-Ras) were confined to domains roughly 200 nm in size. In this instance, cholesterol depletion did not affect confinement. For H-Ras, the mobile fraction was 73% with a diffusion coefficient of 0.53 μm2/ s (SPT) or 0.48 μm2/ s (FRAP) (42). Interestingly, the 200 nm confinement of the slow-diffusing fraction was only observed for active (GTP-bound, ‘raft-associated’) Ras, not the inactive (GDP-bound) form (observed by both mutagenesis and insulin activation of H-Ras) suggesting that confinement is not solely controlled by lipid–lipid interactions (19,43). Similar confinement has been reported for a G-protein coupled odorant receptor with ∼ 50% of the receptor being confined to larger domains (300–550 nm), ∼ 30% in small domains (180–200 nm), and the remaining 20% being either immobile or diffusing freely (44). For a different G-protein-coupled receptor, the μ-opioid receptor, Daumas et alfound that 90% of the receptors diffuse within a domain that diffuses itself, a motion they termed as walking confined diffusion (45). Taken together, these studies suggest that confinement areas are unique to each of the observed receptors, in turn suggesting that specific receptor interactions define these zones rather than generic ‘raft domains’.

Similar to the size of the observation area in FRAP, the sampling frequency in SPT experiments can also influence values of the diffusion coefficients and data interpretation. By developing SPT with ultra-high temporal resolution (25 μs instead of the typical video rate of 30 ms), Aki Kusumi was able to detect different type of diffusion, termed ‘hop diffusion’, in which the entire membrane is compartmentalised into 30–250 nm-sized compartments (6,46,47). Proteins and lipids are slowed down as they hop from one compartment to the next, but within each compartment diffusion is not significantly lower than that observed for free diffusion. Hence, these membrane compartments are distinctly different to TCZs formed by lipid microdomains (48) and are thought to be the result of an underlying membrane skeleton to which transmembrane proteins are anchored creating a ‘picket-fence’ within the membrane (49). It should be remembered that to distinguish between Brownian motion and hop diffusion requires a time resolution of tens of microseconds, without which hop diffusion is simply interpreted as free diffusion with a slower average diffusion coefficient.

Multi-Parameter Imaging: Combining Diffusion With Concentration

A key finding of raft characteristics by the theoretical calculation by Nicolau et alis that global diffusion retardation by high viscosity islands increases the rate of protein–protein interactions (30). Further, the existence of raft domains with high viscosity can increase the local concentration of proteins that have a low average density but high raft-affinity on account of their specific membrane anchors. Maximal collision rate, a measure of signalling efficiency, is achieved when 10% of the membrane is covered by small (∼ 6 nm diameter), mobile domains and Draft/Dnon - raft = 0.5 (values of Draft/Dnon - raft = 0.3–0.5 have previously been reported in cells (35,41)). Raft domains of that nature also modestly increase the collision rate of proteins even if they have no affinity for the raft domain. The authors point out that this may have important biological consequences for different types of membrane domains, e.g. caveolae (large and immobile) and GPI-protein-enriched domains (small and mobile). However, it is predicted that experiments require a time resolution of 5–10 ms to detect these small raft domains by SPT, which is right at the technical limit of most set-ups.

An alternative single molecule technique is FCS (Figure 4), first demonstrated by Elson in 1974 as a method for analysing molecular kinetics (50). In its most basic form, the femto-litre illumination volume of a laser scanning confocal microscope is held stationary. The intensity fluctuations as a low concentration of fluorescent molecules diffuse in and out of this spot are then recorded using high-speed detectors (51). From this time-series of intensity measurements, an auto-correlation analysis allows molecular diffusion rates to be obtained with high statistical accuracy and over a wide range of time scales—modern detectors achieve a range from microseconds up to seconds (52). The method uses low laser powers and low probe concentrations making it applicable to live cell measurements with minimised perturbations to the membrane. Recently, several related techniques have been developed and continue to be developed. These include, for example, two-colour cross-correlation spectroscopy (FCCS) (53), image correlation spectroscopy (ICS) (54) and raster-scanned FCS (raster ICS or RICS) (52). These new methods have the ability to map the spatial distribution of diffusion coefficients at selected membrane regions, as demonstrated for the mobility of paxillin at focal adhesions (55).

Figure 4.

FCS. A) As single molecules diffuse into the stationary spot illuminated by a confocal microscope, bursts of fluorescence are detected. Autocorrelation analysis of this burst pattern can give the transit time of a molecule through the spot by comparing a half-value time τcorr at amplitude (a/2) relative to the initial correlation amplitude a. B) Variable spot-size FCS can distinguish between a protein meshwork (blue), free diffusion (green), and dynamic partitioning (red) even for sub-resolution membrane structure (shaded area) by measuring the intercept with the transit time axis.

The advantage of FCS is that it allows one not only to extract diffusion coefficients and identify anomalous diffusion but that it also measures the number of molecules in the observed spot. For example, fluorescence fluctuation analysis has been used to determine the degree of clustering of the EGF receptor (22) and the stoichiometry of protein complexes (56). Thus, FCS has developed into a truly multi-parameter imaging mode that has been used to describe the dynamics and oligomerisation of GPI-anchored receptors (57), adhesion receptors and tetraspanins (14).

To unravel the relative contributions of membrane lipid microdomains and the actin cytoskeleton, Hai-Tao He and Didier Marguet used variable spot-size FCS and defined an ‘FCS diffusion law’ (Figure 4B), which extracts structural information from below the diffraction limit of optical microscopy (58). Microdomains and cytoskeletal confinement impart different deviations on the relationship between the diameter of the observation spot and transit time that a molecule takes to diffuse through that spot (59). On a plot of diffusion time against area, dynamic partitioning results in a positive intercept with the y-axis whereas a meshwork produces a negative intercept (unhindered diffusion passes through the origin) (58). This was shown theoretically and confirmed with observations of Thy-1 (raft associated—positive intercept) and transferrin receptor (transmembrane protein hindered by the cytoskeleton—negative intercept). This approach has since been used to demonstrate that the protein kinase Akt dynamically partitions into microdomains in cells via its pleckstrin homology (PH) domain (23).

FCS has now been combined with total internal reflection (TIR-FCS) excitation (60,61). Such a set-up allows the measurement of the local probe concentration and the local translational mobility within the membranes as well as kinetic rate constants that describe the association and dissociation of ligands with membrane receptors. Ohsugi et aldemonstrate the improved accuracy of TIR-FCS over FCS by imaging a farnesylated variant of green fluorescent protein (GFP) (20). While SPT under TIR illumination cannot easily track cytosolic proteins due to their 3-D motion, TIR-FCS was able to quantify the proportion of the GFP construct that was membrane bound. Further technical improvements for FCS measurements specific to membranes are summarised in a recent review (62).

An exciting prospect is the integration of time-resolved single photon counting (TCSPC) techniques with unrestricted data acquisition to store spatial, temporal, spectral, and intensity information (63). This allows the correlation of dependencies between various fluorescence parameters. For example, FCS autocorrelation curves can be constructed for fluorescent events of a certain fluorescence lifetime. This can not only help to remove background and autofluorescence and thus makes diffusion measurements more accurate, but also enables a single experiment to give information about the kinetics of, for example, monomers and dimers, local concentrations of proteins and their different states, their relative stoichiometry, and affinities. By combining this information with mathematical models, we are in a position to take a big step forward towards a detailed understanding how proteins function in the cell membrane.

Conclusions and Perspectives

The recent technical advances in fluorescence microscopy hardware and analysis have enabled researchers to quantify the dynamics of membrane proteins in live cells. FRAP, FCS, and SPT go beyond traditional fluorescence imaging and have converted microscopes into molecular measurement tools that can operate on many spatial and temporal length scales. Synergistically, these approaches have helped develop a picture where the diffusion characteristics of membrane proteins show complex and subtle behaviour. The influence of the type of the membrane association on protein diffusion; diffusion restriction imparted by the membrane cytoskeleton and anchored-protein fences; protein oligomerisation, clustering, and protein–protein interaction; and the dynamic partitioning into small, heterogeneous, and mobile lipid microdomains have been measured with remarkable accuracy given the resolution limits of optical microscopy. The arrival of super-resolution microscopy approaches such as stimulated emission depletion (STED), photo-activated localisation microscopy (PALM), and stochastic optical reconstruction microscopy (STORM) have further improved the accuracy of dynamic measurements, as recently demonstrated by combining PALM with SPT (64) and STED with FCS (65). The reduced observation volume of STED detected differences between the FCS autocorrelation curve of phosphoethanolamine and that of sphingomyelin whose diffusion is transiently trapped in cholesterol-enriched complexes (65).

The interdependence of membrane organisation and protein diffusion is, however, still poorly understood. Furthermore, current models do not adequately take into account active transport systems such as directed flow, membrane budding, recycling, and so forth. Answers to these questions may come from multi-dimensional microscopy techniques that have the ability to extract more detailed information about molecular diffusion rates and other membrane parameters. So far, we have a limited understanding of how local variations in membrane fluidity affect protein concentration, oligomerisation, and diffusion. We may be able to address this question by integrating FCS or SPT with the imaging of membrane lipid order using environmentally sensitive probes such as Laurdan (66) or di-4-ANEPPDHQ (67). Quantitative multi-parameter microscopy could further hold the key to delineate how global cellular manipulations and treatments act on the localised and molecular organisation of membrane domains and compartments, which in turn influence protein dynamics, interactions, and functions.

Acknowledgements

We would like to thank the Australian Research Council and the National Health & Medical Research Council of Australia for funding.

Ancillary