The cellular membrane is the primary theater of signal transduction, which begins at the receptor binding at the membrane surface (1, 2). Transient protein-protein and proteinlipid interactions on membrane surfaces constitute the key steps in these signaling cascade reactions (2-5). These interactions are often combinatorial in design, involving reactions such as homo- and hetero-oligomerization (6-10). However, the molecular details of many such systems still remain mysterious. This is largely because mechanistic investigations, which must constitute quantitative measurements, on self-assembled lipid bilayer structures can be a prodigious technical challenge. On experimental systems utilizing live cells, there are added complications due to the copious number of additional background processes which may or may not participate the process under observation. Fortunately, many advances in membrane reconstitution as well as imaging and spectroscopic techniques have made it possible to probe quantitative information such as kinetic rate constants of membrane association and dissociation, catalytic rates, and diffusion constants (11-14). In particular, supported lipid bilayers (SLBs) stand out in its straightforwardness in preparation, experimental implementations, and data analysis (15-21). SLBs are planar lipid bilayers that can be as large as square centimeters in area. They are typically prepared by the rupturing small unilamellar vesicles, which then fuse with each other into a single bilayer on a solid support such as glass coverslips (16, 22).
One way to detect protein-protein interactions on membrane surfaces is by measuring diffusion, which is sensitive to the size of the proteins as well as their conformations, orientations, and overall interaction with the membrane. In addition to the obvious advantages of SLBs in allowing precise control over their chemical environment, another important advantage is that the SLB lipids and other molecules embedded within are fluid and exhibit unencumbered Brownian diffusion (17, 23, 24). This is in stark contrast with the membrane diffusion in live cells where diffusion is usually anomalous (24). The well-defined diffusion behavior allows the diffusion measurements performed on SLBs to be interpreted without much ambiguity. Thus, SLBs have been demonstrated in its usefulness in many works examining molecular interactions on membrane surfaces. Two such examples are K-Ras4B GTPase (25) and the pleckstrin homology (PH) domain of Bruton’s tyrosine kinase (Btk) in their dimerization behavior (26). Homodimerization is perhaps the simplest possible protein-protein interaction and is ubiquitous in signal transduction. Nevertheless, the accurate determination of its kinetics is still a significant technical challenge, particularly if it is dependent on membranes. This is the case for both K-Ras4B and Btk, whose dimerization has never been observed in solution even though they were thought to be critical in their function. In this review, the diffusion-based investigation of their dimerization behavior of K-Ras4B and Btk using fluorescence spectroscopy and imaging will be described.
While it can be reasonably assumed that the diffusion coefficient of a molecule tethered to a membrane would be dependent on its overall size, their theoretical relationship is not necessarily straightforward. In a three-dimensional space, particle diffusion follows Stokes-Einstein relation, where the diffusion coefficient
For peripheral membrane proteins, many factors influence the observed diffusion:  the number of lipid anchors attached to the protein, if any,  the degree of immersion and interaction (including nonspecific ones such as Coulomb interactions) of the protein with the membrane,  the structure and orientation of the protein with respect to the membrane, among others (30). Without a detailed knowledge of these factors, it is in principle impossible to predict the diffusion behavior of the peripheral membrane proteins. A useful alternative is to experimentally calibrate the diffusion coefficients as a function of the molecular size of the proteins. In an
Fluorescence spectroscopy and microscopy are by far the most common methods to measure diffusion on biological membranes. Among various techniques, fluorescence correlation spectroscopy (FCS) and single-molecule tracking (SMT) with surface-selective total internal reflection microscopy (TIRFM) offer high precision and reliability and have been used in a number of studies seeking to establish molecular interactions (Fig. 1), including the K-Ras 4B and Btk PH domain studies.
FCS is a type of fluorescence fluctuation spectroscopy (FFS), which are techniques based on the time-dependent fluorescence intensity fluctuations. It was invented in more than thirty years ago (32), but it has not been very useful until the widespread use of very stable lasers in more recent years. FCS is now commonly used to characterize biological molecules in solutions, membranes, and in live cells (33, 34). It relies on observing fluorescent molecules in a laser focus, which is typically close to diffraction-limited and very small focus volume (∼10−15 L). Therefore, the number of molecules observed is very small— which is why sometimes they are considered “single-molecule” techniques (35). The time-dependent intensity fluctuations are due to fluorescent molecules diffusing in and out of the focus, and therefore contain information about the average number of the molecules within the focus volume and their diffusivity (Fig. 1A). To extract these information, the autocorrelation function (ACF) of the fluorescence intensity fluctuation trace is computed, and it is fit to a autocorrelation model with Gaussian focus G(
An added benefit in using FCS to detect multimerization of fluorescent molecules is that it is particularly sensitive to brighter species. If there are mixed species present in FCS, the autocorrelation model is given by (Equation 4)
In SMT, the motions of single fluorophores are directly visualized and captured by a high quantum efficiency camera (4, 23, 30, 36-38). By using extrinsic fluorophores of high quantum yield and photostability, long-lived, high signal-to-noise ratio trajectories of single molecules can be recorded (Fig. 1B). In the data analysis, the molecule position is identified, and by linking them the particles between teach time frame, a trajectory for each molecule is constructed (39). Thousands of such trajectories can be collected in a single movie, and the diffusion coefficient
K-Ras4B is one of four isoforms of Ras, which is a central molecular switch in a number of signaling pathways such as the mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinases (PI3K)-AKT pathways (42, 43). A small globular protein that is tethered to the membrane by a lipid anchor, Ras—and especially K-Ras4B—is of a paramount clinical importance in that its oncogenic mutations, where it is unable to return to the GDP-bound “off” state, are found in ∼30% of all human cancers (44). Despite concerted efforts to develop therapeutic strategies to inhibit Ras, all conventional, small molecule-based competitive inhibitors to date have failed (45). This has led to an interest in novel strategies to target Ras by modulating its spatial organization, as Ras is observed in dimers and clusters on cellular surfaces (46-49). In particular, it was speculated that its dimerization drove the assembly—this was supported by the consistent dimeric structures in crystals, as well as the fact that some interacting partners of Ras do so in pairs (for example, SOS inter-acts with two Ras molecules) (50-52). However, Ras dimers are never observed in solution, and despite many attempts to detect them on membranes no consensus exists regarding the nature of Ras dimers—whether they are intrinsic properties of Ras, or if so, which specific interactions are responsible—and even whether they exist at all (41, 53-56).
Whether natively expressed and fully processed K-Ras4B can homodimerize on its own on membrane surfaces without extrinsic factors was investigated by using SLB-based reconstitution, FCS, and SMT (25, 57). Two fluorescent labeling strategies were used: eGFP fused to K-Ras4B and an extrinsically labeled guanosine nucleotides that binds to K-Ras4B, and both displayed identical results. A definitive answer to this question would involve a measurement of the two-dimensional dissociation constant,
This strategy requires the knowledge of protein surface density and the diffusion coefficients of the monomers and dimers. While the surface density and diffusion coefficient of monomers may be readily measured by FCS or SMT, it is less straightforward to measure the diffusion coefficient of dimers of unknown dissociation constant, especially if it is weak. This problem was addressed by deliberately inducing dimerization using a strong cross-linker, in this case the Ras binding domain (RBD) of c-Raf fused to leucine zipper (LeuZ) motif (RBD-LeuZ). RBD selectively binds to GTP-bound Ras, and LeuZ is a constitutively dimeric motif. Therefore, the introduction of LeuZ to GTP-bound Ras results in crosslinking of Ras via LeuZ, and the diffusion coefficient of the dimeric Ras could be determined. In this case, the diffusion coefficient a dimer was approximately half that of a monomer, akin to the system constructed by PH domain multimers (31).
Because it was expected that Ras would have a very weak dimerization affinity, it was important to establish the detection limit of the measurement. Using the experimentally determined values of the diffusion coefficients of monomers and dimers, theoretical FCS titration measurements for a range of dissociation constants were simulated (Fig. 1C, right). The simulation shows that, due to the nonlinear contribution of brighter species to the overall measurement, FCS provides a wide dynamic range from strong to very weak dimerization reactions. The brightness effect is so prominent that even very weak dimerization reactions with
The diffusion measurements indicated that, by itself, K-Ras4B remained unambiguously monomeric. Fig. 2A shows representative FCS and SMT step-size distribution data. In FCS (Fig. 2A, left), GTP-bound K-Ras4B does not change significantly with respect to increasing surface density, unlike K-Ras4B crosslinked by RBD-LeuZ. The estimated
Bruton’s tyrosine kinase (Btk) is a protein kinase that is critical in activation of B-cells (59, 60). Conventionally, its activation depends on the recruitment to the plasma membrane by interaction with phosphatidylinositol (3, 4, 5)-trisphosphate (PIP3) lipids, which are produced upon receptor stimulation. The pleckstrin homology domain fused with a tec homology domain (PH-TH) of Btk selectively binds two PIP3 headgroups, and results in activation of Btk (61-64). Crystal structures show a consistent dimer structures of the PH-TH domain, and evidence suggested that
The membrane-dependent dimerization was studied with FCS on supported lipid bilayers. The eGFP-fused PH-TH domain of Btk (an alternative, cysteine-labeled Alexa Fluor 647 also resulted in identical behavior) was introduced to SLBs containing a trace amount (0.005%) of fluorescently labeled lipids (Texas Red-DHPE) and 1-4% PIP3, and which recruited the protein to the membrane. The diffusion of both eGFP-Btk PH-TH domain and Texas Red-DHPE were simultaneously measured by two-color FCS. The titration curve was obtained by plotting the relative diffusion (PH-TH diffusion divided by lipid diffusion) as a function of the protein surface density. This had an effect of correcting for variations in the measured diffusion arising from microscopic heterogeneity in the membrane environment.
The left panel of Fig. 2B shows the FCS titration measurements for the wild type Btk PH-TH domain. There is a robust decrease in the overall diffusion as the surface density is increased on a 4% PIP3 membrane, consistent with dimerization. It was also found that when the PIP3 density lower at 1%, there is a substantially less change in diffusion, indicating smaller fraction of dimers at the same protein surface density. To confirm that this observed change in diffusion is due to dimerization, the same experiment was performed with a mutant construct whose putative dimerization interactions (65) had been removed. The resulting titration curve (Fig. 2B, right panel) shows that the diffusion is no longer modulated by the increased surface density, indicating that dimerization predicted by the crystal structure is responsible for the observed changes in diffusion. This observation, combined with adsorption kinetics measurements by total internal fluorescence microscopy, led to a Btk activation model in which each Btk PH-TH domain interacts with two PIP3 lipids then dimerizes and activates through
Dimerization and higher order multimerization of membrane proteins are a recurring motif in the cellular signal transduction, and a possible protein-protein interaction target for therapeutic interventions. However, it has been challenging to observe complexes and assemblies with traditional
The authors have no conflicting interests.