Proteins within most macromolecular complexes or organelles continuously turn over. This turnover results from association and dissociation reactions that are mediated by each of the protein's functional domains. Thus, studying organelle or macromolecular formation from the bottom up using theoretical and computational modeling approaches will necessitate the determination of all of these reaction rates in vivo. Yet current methods for examining protein dynamics either necessitate highly specialized equipment or limit themselves to basic measurements. In this protocol, we describe a broadly applicable method based on fluorescence recovery after photobleaching (FRAP) for determining how many reaction processes participate in the turnover of any given protein of interest, for characterizing their apparent association and dissociation rates, and for determining their relative importance in the turnover of the overall protein population. Experiments were performed in melanoma M2 cells expressing mutant forms of ezrin that provide a link between the plasma membrane and the cortical actin cytoskeleton. We also describe a general strategy for the identification of the protein domains that mediate each of the identified turnover processes. Our protocol uses widely available laser-scanning confocal microscopes, open-source software, graphing software and common molecular biology techniques. The entire FRAP experiment preparation, data acquisition and analysis require 3–4 d.
Protein turnover and reaction-diffusion processes in cells
Proteins within most macromolecular complexes in the cell continuously turn over. Gaining a quantitative understanding of how macromolecular structures are assembled and how long they persist will ultimately necessitate the systematic determination of the kinetics of association and dissociation of each participating protein. To add further complexity to this picture, proteins usually possess several distinct functional domains, each mediating interactions with other proteins, which may potentially lead to changes in protein conformation. Thus, the reaction kinetics of all of the protein domains need to be characterized in each relevant condition. In vitro biochemical assays enable the accurate determination of reaction rate constants of purified proteins in dilute buffers. However, in the crowded cellular environment, reaction rates can be modulated by complex signaling networks that may not have been exhaustively described. Consequently, new methods for characterizing apparent association and dissociation rates within living cells are necessary to help us understand macromolecular complex formation from the bottom up.
Within cells, protein components of large macromolecular complexes (such as the cytoskeleton, organelles or the nucleus) are present in two pools: a pool bound to the macromolecular structure and a pool that can diffuse in the cytoplasm. Thus, turnover of a protein comprises both a reactive component, which results from all of the binding reactions through which it can associate with the macromolecular complex, and a diffusive component, which is influenced by the organization of the cytoplasm and/or the structure to which the proteins are associated. If diffusion and reaction act on very different timescales, it is possible to distinguish their contributions to turnover using simple analyses ('uncoupled reaction and diffusion'), whereas if they occur on similar timescales more complex strategies based on modeling are necessary to decouple reaction from diffusion1,2. This protocol will restrict itself to uncoupled reaction-diffusion, and it is broadly applicable to any protein that forms part of any large cellular organelle.
The overall kinetics of turnover of a protein depend on the reaction rate constants of each of the domains that have a role in its localization, as well as on the portion of the total protein population bound by one or several of these domains to the structure to be studied. Changes in the turnover kinetics of a protein can be due to changes in the reaction rate constants of one or several of its domains and these molecular-level changes can lead to profound cellular-scale changes. Although we have used the cytoskeleton as an illustration of this protocol, it is of course not limited to the study of cytoskeletal proteins.
Methods for characterizing protein turnover
In living cells, protein kinetics has been probed with a variety of techniques, including FRAP, fluorescence loss after photoactivation (FLAP), fluorescence correlation spectroscopy (FCS) and single-molecule imaging. Here we briefly describe the principle, advantages and drawbacks of each technique.
FRAP is perhaps the most commonly used technique for examining protein turnover in cells, owing to the simplicity of its implementation on most laser-scanning confocal microscopes (Fig. 1a). In FRAP experiments, a small region of interest (ROI) is bleached by a short exposure to high-power laser light, and then fluorescence recovery is monitored over time. Experiments reveal whether a protein is mobile or not. The results are generally quantified by measuring a recovery half-time and the fraction of protein that appears immobile on the timescale over which recovery was followed (Fig. 1a,b). This is sufficient to reveal gross differences in behavior, such as changes in response to silencing of proteins3. However, the shape of the FRAP recovery curve reflects all of the complexity of the dynamics of the protein of interest, and as a consequence more precise information can be extracted from it. For most proteins within macromolecular complexes, recovery stems from both diffusive (either in an organelle or in the cytoplasm) and reactive processes. The presence of a substantial immobile fraction may be the result of excessive imaging-induced photobleaching or depletion of a major fraction of the total fluorescently tagged proteins from the organelle to be studied (this is the case in particular in small cells such as yeast4), or it may signify that recovery has been followed over a duration that is short in comparison with the protein's actual recovery time. In most cases, more quantitative analyses are necessary to understand the molecular origin of recovery kinetics. Indeed, when the fraction of freely diffusing protein is large compared with the fraction of bound protein, changes in reaction kinetics may not be resolvable. For pure diffusion, fluorescence recovery curves can be fitted with analytical solutions of diffusive recovery5,6,7,8 that can be corrected for the limited acquisition rate of the instrument9,10. When both diffusive and reactive processes take place, careful examination of the recovery curves often reveals that multiple processes participate in recovery11,12,13,14,15. Diffusive processes can be experimentally separated from reactive processes by examining recovery in regions of different radii. Indeed, the characteristic timescale τ of 2D diffusive processes scales as τ ≈ R2/4D, where R is the radius of the region and D is the diffusion constant of the protein5,6,7. In contrast, the timescale of reactive processes is insensitive to R. When diffusion operates on a very different timescale to that of the reaction, the interplay between reaction and diffusion can be separated ('uncoupled reaction and diffusion'), enabling the estimation of the diffusion constant, as well as the association and dissociation rates13,16. For purely diffusive processes and circular bleach regions, fluorescence recovers at a rate proportional to 1−t/τ, where τ ≈ r2/(γD) is the characteristic diffusion time and γ is a constant that depends on dimensionality (γ = 2 for 1D diffusion, γ = 4 for 2D diffusion and γ = 6 for 3D diffusion)6. However, in the presence of nonzero flux boundary conditions, the dimensionality of the fluorescence recovery curves on membranes (in 2D) or in the cytoplasm (in 3D) may differ and cannot be approximated by a power law dependency17,18. Moreover, for the FRAP associated with more complex diffusive behaviors, such as anomalous diffusion, we refer interested readers to refs. 19,20. In contrast, for first-order reactive processes, fluorescence should recover as e−kofft, where koff is the dissociation rate constant. However, the identification of the molecular origin of each recovery process is often challenging. Generally, numerical simulations of both diffusion and reaction systems allow more rigorous consideration of the experimental setup and the specific boundary conditions of the FRAP experiment, and they allow exhaustive hypothesis testing but are specific to each particular protein to be studied.
Other techniques monitor the loss of fluorescence occurring in a small region of an otherwise dark cell. This experimental condition can be achieved either by photoactivating a small region (FLAP)13,14,21 or by photobleaching the whole cell except for one small region11,22 (FLIP: fluorescence loss in photobleaching). In both conditions, fluorescence decay kinetics is usually dominated by reactive processes. Careful examination of the decay curve can enable identification of the multiple processes contributing to fluorescence loss11,13,14. One drawback of this implementation is that it is very sensitive to imaging-induced loss of fluorescence. Although the methods presented in the current protocol are for FRAP, they are also applicable to FLAP with minor modifications14.
FCS is used to measure protein concentrations, protein diffusion constants and the kinetics of complex formation in vitro23. To do so, FCS monitors fluctuations in fluorescence intensity within a diffraction-limited volume that arises from the diffusion of fluorescently tagged proteins into and out of this control volume. Fluctuations in the fluorescence signal are then analyzed by autocorrelation analysis. This yields a curve that relates the correlation of the fluorescence signal to itself as a function of time delay. At short times, signals are strongly autocorrelated but at long times they are not because most of the proteins have diffused out of the control area. Determination of aggregation relies on the measurement of diffusion constants using appropriate analytical models. Although it is used extensively in vitro with purified proteins, its application to cells is more challenging because the analytical models of diffusion are less accurate owing to the complexity of the intracellular environment compared with that of dilute solutions (including steric hindrance, active transport), and because of the plethora of reactions that individual proteins can undergo in the cell24. Finally, the determination of the protein association and dissociation rates with membranous or cytoskeletal structures remains complex. Instrumentation for FCS is less widely available than FRAP instrumentation, although variants of FCS can be implemented on spinning-disk confocal microscopes with high-performance electron-multiplying charge-coupled device (EM-CCD) cameras25 or on laser-scanning confocal microscopes26.
Single-molecule imaging enables direct visualization of protein attachment and detachment from immobile subcellular structures27. In single-molecule imaging experiments, cells are transfected or microinjected with very low numbers of fluorescently tagged proteins and imaged with sensitive cameras28,29,30. Contrast formation results from the low concentration of the fluorescently tagged proteins and the very long image acquisition times used (>500 ms): proteins that diffuse in the cytoplasm contribute to background signal, whereas proteins that stay immobile during the time frame of acquisition (either because they are bound to an immobile structure or because they diffuse very slowly) appear as diffraction-limited fluorescent speckles28. For quantitative measurements of protein dynamics, it must be possible to distinguish and track individual speckles. In particular, plotting the speckle lifetime distribution allows the determination of the dissociation constant koff of the protein. Thus, single-protein imaging necessitates ultrasensitive cameras and accurate speckle-tracking programs29. One drawback is that the long exposures necessary to achieve sufficient signal noise for tracking mean that only few images can be acquired before imaging-induced photobleaching becomes a serious issue.
Here we describe a broadly applicable protocol based on FRAP, which allows quantitative determination of the association and dissociation rates of each of a protein's subdomains within living cells using widely available instrumentation and minimally complex analysis14,15. Our protocol relies on the separation of the contribution of diffusive processes from reactive processes in the fluorescence recovery after photobleaching. The number of molecular processes that participate in reactive recovery is determined by fitting the experimental fluorescence recovery curve with an increasing number of first-order processes. The molecular nature of the reactive recovery processes can then be identified by separately measuring the recovery kinetics of each of the subdomains of the protein of interest and comparing these with recovery of the full-length protein. In addition, the proportion of total protein participating in each turnover process can be determined. Our protocol uses widely available confocal scanning laser microscopes to acquire experimental data and common software to determine the number of processes participating in recovery. Our strategy for matching subdomains with first-order processes is straightforward, and it makes use of common cloning techniques. Furthermore, our protocol is generally applicable to any experimental situation in which proteins bind to a membrane, the cytoskeleton or any cellular organelle.
Application of this protocol
Our technique can be used to investigate the molecular processes leading to protein turnover with FRAP. In a previous study14, we examined the turnover kinetics of the actin-binding protein α-actinin, which dimerizes to cross-link actin filaments. We found that its reactive recovery comprised two separate molecular processes: one fast process accounting for two-thirds of the recovery and one slow process accounting for one-third of the recovery. By comparing the recovery kinetics of the α-actinin–binding domain alone with that of the full-length protein, we showed that the fast process was due to α-actinin dimers binding to only one actin filament and, hence, that it did not have a role in cross-linking. Therefore, in studying the mechanics of actin networks cross-linked by α-actinin, one must consider the second, slower timescale due to dimers rather than the first, faster and dominant timescale. In a second study, we examined the turnover dynamics of the membrane F-actin linker protein ezrin15. This revealed that ezrin's association with the actin cytoskeleton was very labile, whereas its association with the cell membrane was much more stable. We anticipate that our protocol can be used to discern the role of each of a protein's subdomains in dictating full-length protein turnover and localization. This will pave the way for a quantitative understanding of how protein dynamics alter the organelle mechanical properties.
Most proteins possess multiple domains that allow them to bind to interactors, and each of these molecular interactions participates in setting the dynamics of turnover. Overall, protein turnover as measured by its half-time depends on the fraction of bound/unbound protein, the total number of possible molecular interactions, the proportion of protein undergoing each interaction and the reaction rate of each interaction. Each of these variables can be modified by signaling, such as an increase in the proportion of phosphorylated protein in response to progression through the cell cycle. Post-translational modifications can result in changes in protein structure and consequently lead to changes in the binding affinity of some or all of a protein's subdomains or the exposure of cryptic binding domains. If a sufficient portion of the total protein is affected by these post-translational modifications, this will lead to changes in localization and recovery half-time. However, sole measurement of the half-time t1/2 cannot identify the origin of changes in recovery rate. Indeed, changes in t1/2 may result from changes in the fraction of bound/unbound protein, changes in the total number of reaction processes i participating to recovery, changes in the characteristic times τd,i of some or all of the reaction processes, changes in the relative importance fi of some or all of the reaction processes or a combination of all of these factors. By allowing the determination of this information, we anticipate that our protocol will enable a more precise understanding of the molecular origins of changes in turnover and protein localization.
Computational models of macromolecular assemblies have become an integral part of research into the organization and mechanics of the cytoskeleton during cellular processes such as cell migration31,32,33, cytokinesis34,35 or spindle organization36. These models attempt to test our understanding of cellular processes from the bottom up by incorporating the biochemical reactions known to participate in the process to predict macroscopic properties that can be measured experimentally. Thus, measurement of the different biochemical reaction rates that participate in protein localization is necessary as an input for such simulations. We anticipate that our technique will complement these computational approaches by experimentally characterizing the necessary biochemical reaction rates and testing model predictions.
Limitations of the analysis
Interpreting protein turnover dynamics in terms of first-order reaction kinetics is naturally limited to linear assembly kinetics, and it only approximates more complex binding reactions such as cooperative binding, as well as more complex diffusion kinetics such as hopping diffusion or protein-trapped diffusion37. However, such limitations are a fundamental problem of FRAP experiments. Furthermore, FRAP experiments cannot provide information about cooperative interaction dynamics between the two different molecules. If such phenomena are at play, FCS measurements will be necessary38.
Another limitation of our analytical approach is that we can only distinguish molecular processes that occur on sufficiently different timescales. If several molecular processes occur at similar timescales, they cannot be separated, and the apparent rate constant measured reflects an average over all of the molecular processes acting at that timescale. In that case, the apparent rate constant results from multiple molecular processes and does not represent a real molecular binding rate. In practice, this is the case if fewer turnover processes are identified than domains mediating association with the macromolecular structure of interest. In these situations, interpretation of the results will necessitate in-depth consideration of all of the reactions that can lead to fluorescence recovery, which is something best achieved by using computational modeling approaches.
Accurate separation of the diffusive processes from reactive processes is crucial for our protocol. Failure to correctly identify the diffusive processes that contribute to recovery along with their timescales may lead to incorrect identification of the number of participating reactive processes. Indeed, diffusive recovery takes the form of a power law, and increasing the number of exponentials used to fit fluorescence recovery may lead to a progressively better fit, leading to erroneous conclusions.
Finally, the presence of convective flows, such as those driven by actomyosin contractility21,39,40, may confound analysis. In this case, it may be necessary to determine convective flow velocity using co-transfected markers in each experiment to enable correct interpretation of the FRAP data.
To characterize the apparent reaction rates participating in the turnover of proteins bound to large and relatively immobile subcellular structures, we devised an integrated experimental and analysis protocol14,15. Our protocol involves classic photobleaching experiments with a laser-scanning confocal microscope (Fig. 1a), but it carries out a more in-depth analysis to extract information about protein association and dissociation rates. To identify the origin of the molecular processes participating in turnover, the protocol can be combined with classic molecular biology techniques to generate constructs that encode the protein's domains with fluorescent tags and to compare their recovery with the full-length protein (however, these steps are not described in this protocol; see below). Thus, our approach can be subdivided into three main stages: the characterization of the timescale of diffusive and reactive recovery, the determination of the number of reactive processes participating in recovery and the identification of the molecular origin of each reactive process.
The first stage in the experimental process is the determination of the timescale of diffusive recovery and reactive recovery in the subcellular structure of interest. Generally, cytoplasmic diffusive recovery operates with diffusion constants on the order of ∼20 μm2/s (Fig. 2), indicating that it is many-fold faster than reactive recovery and is often complete by the first postbleaching frame. Diffusion of proteins in the plane of membranes is considerably slower than in the cytoplasm, with typical diffusion constants on the order of ∼0.1–1 μm2/s (Fig. 2f). As a consequence, it is usually possible to separate reactive recovery from diffusive recovery on the basis of timescales. However, when diffusive and reactive recovery processes operate on comparable timescales, it is not possible to determine reaction rates without the use of complex numerical simulations1,2. Experimentally, diffusive recovery can be identified by acquiring FRAP experiments for a range of radii R of the bleach region (Fig. 2e). Indeed, the timescale of diffusive processes will scale as R2, whereas reactive processes should be insensitive to it. Alternatively, additional techniques such as single-molecule imaging can be used to confirm the nature of the relevant processes.
In the second stage, the portion of fluorescence recovery due to association/dissociation reactions is fitted with functions composed of an increasing number of exponentials of the form Fi(t) ≈ 1–exp(−t/τd,i)fiF0, where F0 is the initial fluorescence of the bleached region and i is a molecular process participating in recovery. Each function Fi represents the contribution of the molecular process i to the total recovery, where fi is the portion of the total protein population undergoing the turnover process i (Σifi = 1), and τd,i is the characteristic dissociation time of process i. The characteristic dissociation time τd is linked to the half-time reported in most FRAP experiments by the relation t1/2 = ln2τ t1/2 = ln(2)τd. The apparent association time τa,i can be calculated as τa,I = (τd,i / fi) × (F0/F ref. 14). If turnover results from association/dissociation of a protein to/from the structure being studied, τd can also be expressed as an apparent dissociation rate ωd = 1/τd.
Once the number of reactive processes involved in turnover has been determined, the third stage is the identification of their molecular origin. As protein domains can usually function independently of one another, the turnover kinetics of the full-length construct should be reflected by the turnover kinetics of its subdomains. The protein subdomains that are likely to have a role in turnover can be identified from previous studies or from databases such as SMART (http://smart.embl-heidelberg.de/), UniProt (http://www.uniprot.org/) or PFam (http://pfam.xfam.org/). Protein interactors can be identified through databases such as String (http://string-db.org/). Subdomains that are likely to have a role in protein localization to the structure of interest can then be fluorescently tagged using classic molecular cloning techniques, as described in ref. 41. The kinetics of each subdomain can then be characterized by FRAP using the same procedures as for the full-length protein14,15 and compared with the kinetics of the full-length protein. Occasionally, more recovery processes are identified than domains interacting with the structure of interest. For example, these supernumerary molecular processes can result from the stabilization of the protein localization owing to interaction via more than one domain, as in the case of proteins containing two actin-binding domains14. Here we underline crucial experimental considerations for carrying out these experiments.
Photobleaching setup and calibration (Steps 1–10).
FRAP experiments necessitate empirical determination of three parameters: the laser settings (to reliably photobleach all fluorophores at a given location), the dimensions of the ROI (to accurately characterize the dynamics of the protein of interest) and the fluorescence loss induced by imaging. The first control needs to be done for each different fluorophore, whereas the second and third must be performed for each different protein examined.
The first calibration experiment seeks to determine the minimal laser settings to bleach all fluorophores in the ROI. In most conditions, proteins that are not bound to a macromolecular structure such as the cytoskeleton or the membrane diffuse rapidly in the cytoplasm over the timescales used for photobleaching. This results in bleach zones that are apparently larger than the chosen ROI and incomplete photobleaching9. To overcome this issue, cells can be lightly chemically cross-linked with paraformaldehyde, which preserves the fluorescence of fluorescent proteins. Laser settings that result in complete photobleaching in the ROI with the shortest bleach duration can then be determined and used for experiments on live cells. It is important to note that only recovery processes with half-times longer than the bleach pulse duration can be reliably detected in FRAP experiments42.
The second calibration experiment aims to determine appropriate ROI dimensions. Indeed, recovery data from regions that are too small are dominated by noise, whereas regions that are too large only allow sampling of recovery at low frame rates. The choice of ROI can also be crucial in separating diffusive recovery from reactive recovery. The characteristic timescale for diffusive recovery is τdiffusion ≈ r2/γD, where r is the radius of the ROI, D is the diffusion constant of the protein and γ is a constant that is dependent on dimensionality (as detailed above). As for globular proteins the diffusion constant increases only weakly with protein mass (see Stokes–Einstein relationship D = kBT/(6πμa), where a is the radius of the protein, and μ is the viscosity of the solution), the diffusion constant of GFP in the cytoplasm can be used as a first approximation (DGFP ≈ 25 μm2/s). For diffusion in the plane of the membrane, diffusion constants of 0.01–1 μm2/s are typically measured10,15,43,44,45. Under these conditions, the timescale for diffusive recovery for an ROI of radius r ∼1 μm is τcyt ∼40 ms for cytoplasmic diffusion and τmem ≈ 1–100 s for diffusion in the membrane (Fig. 2e). In practice, with high-magnification objectives, ROIs of 1–2 μm diameter offer a good signal-to-noise ratio, fast diffusive recovery in the cytoplasm and fast sampling of reactive recovery. In addition, if diffusion in the plane of the membrane is involved, the radius of the ROI may need to be adjusted empirically to allow clear separation of timescales between reactive and diffusive processes based on the fact that the characteristic timescale for diffusive processes scales as τdiffusion ≈ r2/D (ref. 45). For example, the timescale of membrane diffusion can be increased by increasing the size of the bleach region such that it operates on a timescale that is well separated from the timescale of reactive recovery (Fig. 2e). In addition to diffusive processes, convective processes due to myosin-driven F-actin flows have also been observed, for example, in the lamellipodium of cells39,46 or in the cortex40. Thus, if the protein of interest binds to F-actin, one should ascertain whether such flows may contribute to fluorescence recovery. These calibrations and verifications need to be performed for each new protein, and they are specific to each protein.
For analytical interpretation of experimental FRAP data, the choice of an isotropic geometry of the bleaching volume is essential. Although the dimensions of the ROI in the xy focal plane can be set with the acquisition software, the dimension in the z direction is less well controlled and depends on the numerical aperture (NA) of the objective used, the laser power and the pinhole aperture selected on the confocal microscope. If the optimal laser settings are determined in the first calibration experiment, the extent of the photobleached region in the z direction needs to be determined empirically using lightly fixed cells by acquiring a confocal stack of images after photobleaching (Fig. 2b). The ROI can then be chosen to have a diameter equivalent to the extent of photobleaching in z. With 100× oil-immersion objectives with NAs of 1.4, this extent can be as small as 800 nm, depending on the excitation wavelength14. This calibration needs to be performed at least once for each laser setting, and it should be repeated regularly if laser power appears to fluctuate.
The third calibration involves the determination of the loss of fluorescence due to imaging. This involves acquiring a set of control time-lapses with the same frame rate and total number of frames as were used during the experiments to evaluate the extent of loss of fluorescence that is imputable to imaging alone (Fig. 1b). This control is important because it will indicate when turnover of the protein of interest is complete.
Time-lapse acquisition parameters (Steps 11–14).
Characterization of protein turnover kinetics necessitates empirical determination of the optimal frame rate and total time-lapse duration.
The first stage involves determining the nature of the fluorescence recovery of the protein to be studied. Diffusive recovery can be distinguished from reactive recovery through a simple experimental test: the timescale of diffusive recovery depends on the radius R of the bleached zone τdiffusion ≈ R2/γD (Fig. 2e), whereas the characteristic timescale of reactive recovery does not. Thus, by acquiring FRAP experiments of several regions of different diameters in the preparatory stage, it is possible to distinguish recovery due to diffusion from recovery due to association/dissociation reactions and to determine their respective timescales45. In this protocol, we place ourselves in conditions in which diffusion operates on timescales that are very different from those of reaction, and thus diffusion is decoupled from reaction. If diffusion and reaction take place on similar timescales, then the current protocol is not applicable and more complex approaches are necessary, such as those detailed in ref. 16.
To determine the timescale of cytoplasmic diffusive recovery, an ROI is chosen in a region in which only diffusive recovery is expected (such as the cytoplasm in the case of a membrane-binding protein). As cytoplasmic diffusive recovery is rapid, a short-duration photobleaching experiment monitoring recovery at the highest possible frame rate is performed. Because photobleaching is not instantaneous, diffusive recovery occurs in the interval between the end of photobleaching and the first postbleaching imaging frame. As a result, photobleaching appears incomplete. The timescale of cytoplasmic diffusive recovery is the duration Δtdiff necessary for fluorescence in the ROI to reach a plateau. If diffusion in the plane of the membrane is present, its timescale needs to be determined and recovery should be followed for durations over which the contribution of diffusion in the membrane to recovery is minimal. This can be ensured by adjusting the dimensions of the ROI. Indeed, the dimensions of the ROI can be increased such that the timescale of membrane diffusive recovery (τdiffusion ≈ r2/4D) becomes far greater than the characteristic timescale of reactive recovery (Fig. 2e), thus ensuring that membrane diffusive recovery contributes little over the timescale at which reactive recovery operates.
To accurately identify the number of molecular processes participating in reactive recovery, it is essential to determine the duration over which recovery should be monitored. To this end, a series of FRAP experiments are performed in a region containing the macromolecular structure of interest to determine when fluorescence recovery no longer increases. In subsequent experiments, recovery should be monitored over this duration.
Finally, the optimal frame rate over which to monitor recovery must be chosen. There are several considerations in choosing this rate. First, a frame rate slower than 1/Δtdiff will ensure that cytoplasmic diffusive recovery is complete by the time the first postbleaching frame is acquired and therefore that only reactive processes participate in the subsequent recovery in the absence of diffusion in the membrane (uncoupled reaction and diffusion conditions). This ideal situation is not always possible and, if it is not, cytoplasmic recovery will need to be evaluated in each FRAP experiment. Second, a frame rate such that at least 10–20 data points are acquired during the first 50% of recovery should be chosen to allow accurate identification of fast reactive recovery processes. Third, a total of 50–100 points should be acquired over the whole of recovery to identify slower reactive processes. Although in principle the fastest frame rate should be the most accurate for the identification of recovery processes, in practice, it results in high loss of fluorescence due to imaging. If recovery in the region in which the structure of interest is located is indistinguishable from the recovery in the cytoplasm, this suggests that recovery is purely diffusive. If, in contrast, recovery is significantly slower than in the cytoplasm, reaction processes or diffusion in the plane of the membrane may contribute.
Gathering a coherent set of experimental data (Steps 15–19).
Once calibrations have been performed and the acquisition parameters chosen, experimental data acquisition can begin. In each experiment, an appropriate choice of ROI is crucial for determining turnover dynamics. It must be chosen such that the structure of interest is in its center and does not contain autofluorescent organelles or protein aggregates that will affect the accurate monitoring of fluorescence recovery. If such objects move into the ROI during the time course of recovery, or if the cell moves substantially, the experiment must be discarded. The ROI should also always include the cytoplasm to allow for the estimation of diffusive recovery owing to unbound proteins or confirmation that diffusive recovery is complete by the first postacquisition frame (Fig. 2c). In separate experiments, a control region (the loss ROI, lROI) must be chosen to allow for the evaluation of the extent of loss of fluorescence due to imaging. This region must have the exact same geometry as the ROI, it should be centered on the structure of interest and it should not contain fluorescent aggregates or autofluorescent organelles, but the experiment should be carried out in the same cell (Fig. 2c). Alternatively, if several organelles of interest are contained within the ROI, one of the nonbleached organelles can be used to evaluate the extent of fluorescence loss due to imaging.
Analysis of experimental data (Steps 20–28).
After acquisition, image analysis software such as ImageJ or Fiji is used to extract the fluorescence recovery in the structure of interest, the cytoplasmic fluorescence recovery and the loss of fluorescence induced by imaging. For each experiment, these curves are then plotted in graphing software for examination and fitting. Data are normalized to the arithmetic average of the frames acquired before the photobleaching event, and curves are synchronized such that the photobleaching event takes place at t = 0. First, the loss of fluorescence induced by imaging is examined to verify that it is within the bounds determined by the control experiments. Then, the cytoplasmic recovery is examined. In the case of uncoupled diffusion and reaction, this curve should remain approximately constant over time, with an intensity close to the prebleach intensity. If diffusion and reaction operate on similar timescales, the contribution of diffusive processes in the cytoplasm is subtracted from the recovery in the structure of interest. Next, the reactive component of the fluorescence recovery curve is fitted with an increasing number of exponential functions until break conditions are reached. Finally, all the participating processes are identified by matching their dynamics to the dynamics of the protein subdomains.
To investigate diffusive fluorescence recovery, experimental curves are fitted with functions of the form F(t) = (F(0) + F(∞) × (t/τ)α)/(1 + (t/τ)α), where α = 1 in the case of free Brownian motion in two dimensions6,7,47. Experimental work has shown that the analysis of recovery in 3D can be reduced to the simpler 2D case when low-NA objectives and wide confocal apertures are used48,49. In this protocol, we therefore assume that only 2D normal diffusion takes place, and we refer readers interested in fluorescence recovery in one or three dimensions and/or anomalous diffusion to refs. 20,50. One hallmark of diffusive processes is that their characteristic recovery time depends on the radius R of the bleach zone: fluorescence should recover faster for smaller radii and slower for larger radii, giving the same diffusion constant in both the cases. In contrast, reactive processes do not show such dependence. In sufficiently large cells and organelles, purely diffusive processes should fully recover after photobleaching, and any apparent immobile fraction is due to imaging-induced loss of fluorescence. However, in smaller organelles or cells, a substantial portion of the total fluorophores may have been depleted by photobleaching, leading to incomplete recovery, as is the case in yeast or bacteria4. The imaging-induced loss-of-fluorescence curves are fitted with single exponential functions of the form F(t) = A × exp(−(t−t0)/τ). Their characteristic decay time τ can then, in principle, be used to correct the recovery curves for imaging-induced loss of fluorescence. However, modern confocal systems lead to limited loss of fluorescence and typically do not necessitate the correction of the recovery curves. Furthermore, as both the fluorescence recovery and the imaging-induced loss of fluorescence are nonlinear processes, correction of the recovery curves may lead to artifacts and are thus best avoided. Generally, the timescale of imaging-induced fluorescence loss is one order of magnitude slower than the timescale of fluorescence recovery in the ROI, and as a consequence it has only a small effect on the values derived from analysis of experimental data. In a reaction-diffusion system, reactive recovery is usually substantially slower than cytoplasmic diffusive recovery. Diffusive contributions to recovery can be removed by estimating the diffusive recovery in a small region within the original ROI and subtracting this from the total recovery by adjusting for the relative areas of both subregions (Fig. 2c).
After reaction is separated from diffusion, the fluorescence intensity F(t) of the reactive recovery is fitted with a function comprising a progressively increasing number of single exponential functions i of the form Fi(t) = Ai × (1−exp(−(t−t0)/τi)) with F = ΣFi. The underlying assumption is that all recovery processes result from first-order reactions. More complex reaction kinetics may of course take place, and these can be investigated in combination with computational modeling using open-source packages such as VCell (http://www.nrcam.uchc.edu/), Compucell3D (http://compucell3d.org/), CellSys (http://ms.izbi.uni-leipzig.de/index.php/software/cellsys2010) or E-Cell (http://www.e-cell.org). Initial guesses for the fitting parameters can be inferred from experimental curves. In practice, we first fit with one exponential function, then two and so on (Fig. 3) until the following convergence criteria are reached: the goodness of fit estimated through r2 no longer increases, two processes have exactly the same characteristic time and amplitude, and the sum of squared residuals no longer decreases. After the characteristic times of each process are determined, the dissociation and association rates are computed through ωd,i = 1/τd,i and ωa,i = ωd,i fi F/F0, where F is the total fluorescence in the structure and F0 is that of the cytoplasm. The fractions fi of the total protein subpopulation undergoing each recovery process are given by fi = Ai/F, where Ai stands for the amplitudes derived from multiexponential fitting. The fitting procedure is repeated for each individual FRAP curve, and then all rate constants are used to calculate the mean and s.d. An alternative and model-independent approach to investigating the number of reactive processes participating in recovery is to plot the fluorescence recovery due to reactive processes on a log-lin scale51.
In contrast to diffusive recovery curves, reactive recovery curves may contain an apparent immobile fraction that is not only attributable to imaging-induced loss of fluorescence. If such an immobile fraction is present, recovery has not been monitored for sufficiently long and, as a result, the slowest recovery processes may not be accurately identified. If this is the case, a new set of data may need to be acquired by monitoring recovery over longer durations. One experimental situation that may lead to a substantial apparent immobile fraction is when the rebinding of bleached molecules within the bleached region is not negligible. This is particularly the case if the bleached zone represents a large fraction of the overall volume. We refer interested readers to refs. 39,52, which suggest that this effect will have a marked impact on measurements.
To visualize the different processes participating in the recovery of the full-length protein, a logarithmic acceleration plot can be generated by computing the second derivative of the calculated fits and plotting them on a log-lin scale14,15. These plots consist of piecewise linear segments, with each segment corresponding to a different recovery process. The slopes of these segments are respectively ωd,i. Changes in the recovery rates in response to drug treatment or genetic perturbations result in changes in the slope of these linear segments that can easily be visualized in the logarithmic acceleration plots.
Dissecting the molecular origin of reactive turnover processes.
Whereas the above analysis can determine how many reactive turnover processes contribute to the recovery of the full-length protein, it cannot identify what protein subdomains contribute to each of these. To do this, fluorescently tagged constructs that encode each of the protein's subdomains that may mediate association with the macromolecular structure of interest can be generated using molecular cloning techniques, as described in ref. 41. Next, the steps detailed above can be carried for each domain in turn. The turnover rates of the subdomains can then be compared with the rates of the full-length protein to infer the role of each domain in full-length protein turnover. The inferred identities can then be confirmed by specifically targeting the functionality of each subdomain with genetic or chemical perturbations. For example, the membrane-linker protein ezrin possesses an N-terminal membrane-binding domain and a C-terminal F-actin-binding domain. The dynamics of each of these domains can be studied independently of one another and compared with the full-length protein to reveal which domain contributes to which turnover processes (Fig. 4).
PBS tablets (Fisher Scientific, cat. no. 12821680)
Trypsin-EDTA (1×), 0.05% (wt/vol), 500 ml (Invitrogen, cat. no. 25300-062)
Minimum essential medium (MEM), 500 ml (PAA, cat. no. 11095-080)
FBS, 500 ml (PAA, cat. no. A15-151)
Penicillin-streptomycin (100×), 100 ml (PAA, cat. no. P11-010)
Leibovitz's L-15 medium without phenol red, 500 ml (Invitrogen, cat. no. 21083-027)
Glass coverslips, 25 mm (Thermo Scientific, Menzel Glaeser, cat. no. CB00250RA1)
Culture plates, six well (Sigma-Aldrich, cat. no. CLS3516-1EA)
Lipofectamine 2000 transfection reagent (Invitrogen, cat. no. 11668-019)
Opti-MEM I reduced serum medium, no phenol red (Invitrogen, cat. no. 11058-021)
Target cDNA from a miniprep, 0.5–1 μg/ml (ezrin in pEGFP-N1, Addgene plasmid 20680: pHJ421 (ref. 54); EGFP-N1, Takara-Clontech, cat. no. 632469; ezrin FERM domain, Addgene plasmid 62380 (ref. 15); ezrin actin-binding domain (ABD) in pEGFP-C1, Addgene plasmid 62381 (ref. 15))
Formaldehyde solution, 16% (vol/vol), 10 × 10 ml (Electron Microscopy Sciences, cat. no. 15710)
Formaldehyde is toxic and flammable. Ensure that you wear proper PPE, keep this reagent away from ignition sources and use this reagent within a fume hood.
BSA, 100 g (Sigma-Aldrich, cat. no. A7906-100G)
Pipette tips, 20 and 200 μl (Mettler Toledo, cat. nos. RT-20F and RT-200F)
Pipettors, 20, 200 and 1,000 μl (Gilson, cat. nos. F123600, F123601 and F123602)
Centrifuge tubes, 15 and 50 ml (Corning, cat. nos. CLS430791 and CLS430829)
T25 cell culture flask (BD Falcon, cat. no. 353108)
Eppendorf tubes, 1.5 ml (Treff Lab, cat. no. 96.7811.9.05)
Pasteur pipettes, 3 ml (Scientific Laboratory Supplies, cat. no. PIP4210)
CO2 incubator (Thermo Electron Corporation, Hera Cell 150)
Hemocytometer (Hawksley, cat. no. AC1000)
Centrifuge (Beckman Coulter, Allegra X-12 series)
Biosafety cabinet (Bioquell Microflow, Microflow peroxide class II)
Attofluor coverslip holder (Invitrogen, cat. no. A-7816) or 35-mm-diameter Willco glass-bottom dishes (Willco, GWSB-3522)
Confocal microscope: inverted microscope (Olympus, model IX-81 or comparable) with confocal head (Olympus, model FluoView FV1000 or comparable) with 488-nm laser (20 mW) and photobleaching-capable software
100× oil-immersion objective (NA = 1.4, UPLSAPO100×O, Olympus)
CellMask Green plasma membrane stain (C37608, Life Technologies)
Microscopy software FV10 ASW system software (Olympus)
Graphing software, e.g., OriginLab (http://www.originlab.com)
M2 cell culture
Culture M2 melanoma cells53 at 37 °C in an atmosphere of 5% CO2 in air in MEM supplemented with 1% (vol/vol) penicillin-streptomycin, 1% (vol/vol) glutamine and 10% (vol/vol) FBS. Store the supplemented MEM at 4 °C for up to 4 weeks. Passage the M2 cells when they are confluent, typically every 2–3 d. Split at a 1:5 dilution.
Imaging buffer is Leibovitz's L-15 medium supplemented with 10% (vol/vol) FBS. Store the buffer at 4 °C for up to 2 weeks.
Timing 6–10 h plus 48 h of culture time
Cell growth. Culture M2 melanoma cells under normal growth conditions in a T25 flask until they reach confluence (Reagent Setup).
Trypsinize the cells. Remove the growth medium and replace it with 5 ml of PBS. Remove PBS and replace it with 1 ml of 1% (wt/vol) trypsin-EDTA solution. Return the cells to the incubator for ∼5 min. When they are ready, the cells will detach from the bottom of the culture flask when it is tilted.
Plate cells onto glass coverslips or glass-bottom Petri dishes. Add 4 ml of culture medium to the culture flask to inactivate trypsin. Pipette vigorously to separate the cell clumps. Transfer the cells into a 15-ml Falcon tube. Measure the cell concentration with a hemocytometer. Centrifuge the cells at 200g for 5 min at room temperature (RT; 25 °C). Resuspend the cells to a final concentration of 2 million cells per ml. Place one coverslip in each well of a six-well plate, and then deposit 500,000 cells in the center of each coverslip or directly in the center of the Petri dish (Fig. 2a). Return the samples to the incubator for 45 min to allow cells sufficient time to attach. Add 2 ml of culture medium to each well, and incubate the cells under normal growth conditions (Reagent Setup) for at least 2 h or overnight.
Transfections. Transfect cells on each coverslip with the cDNA encoding the fluorescently tagged protein of interest. For lipofection using Lipofectamine 2000, for each transfection sample, dilute 1 μg of cDNA in 100 μl of Opti-MEM in a 1.5-ml tube and mix it gently. In a separate tube, dilute 2.5 μl of Lipofectamine 2000 in 100 μl of Opti-MEM medium for each sample to be transfected. Mix gently. Incubate the mixture for 5 min at RT. Combine the diluted cDNA with 100 μl of Lipofectamine 2000 diluted in Opti-MEM. Mix gently and incubate the mixture for 30 min at RT to allow complexes to form. For each coverslip to be transfected, remove the growth medium, rinse the cells with 2 ml of PBS and add 800 μl of Opti-MEM. Add the 200 μl of complexes to each well to be transfected. Return the samples to the incubator for 3–4 h. After incubation, remove the Opti-MEM and add 2 ml of culture medium. Incubate the samples overnight.
Do not supplement the Opti-MEM with FCS or antibiotics, as this can interfere with transfection.
The amount of Lipofectamine 2000 needs to be optimized separately for each cell type.
Sample mounting. Mount glass coverslips into Attofluor sample holders, and add 1 ml of imaging buffer (Fig. 2a). Keep the samples in the incubator until needed.
Cover the Attofluor sample holders with a Petri dish lid when in the incubator to avoid excessive evaporation.
Ensure that each glass coverslip is correctly centered within the bottom part of the sample holder before fastening the top part. The coverslip must fit snugly within the 25-mm groove. If necessary, re-position the coverslip by sliding it with the tip of a pair of sharp tweezers.
Calibration sample preparation
Timing 30 min
Sample preparation. Transfect the cells with a protein with a high cytoplasmic fraction such as EGFP-N1 according to Steps 1–5.
Formaldehyde fixation. Wash the wells containing the samples three times with L-15 medium at RT. Remove L-15 medium and replace it with a solution of 4% (vol/vol) formaldehyde diluted in L-15 medium. Incubate the samples at RT for 15 min. Remove the formaldehyde solution and dispose of it appropriately. Wash the samples with 2 ml of PBS three times. Replace the PBS with PBS containing 10 mg/ml BSA and incubate the samples at RT for 10 min. Rinse the cells with PBS. Fixed samples can be kept at 4 °C in the refrigerator for 48 h.
Formaldehyde is toxic and flammable. Perform this step in a fume hood and away from ignition sources. Wear appropriate PPE.
Timing 5–6 h
Setting imaging parameters. All of the acquisition steps are described for an Olympus FV-1000 microscope; however, similar functions should be readily available on most laser-scanning confocal microscopes or FRAP-enabled spinning-disk confocal microscopes. Focus the microscope on a field containing fixed cells. The first step is to determine the adequate imaging parameters for examining the effect of one pulse of photobleaching on the cellular fluorescence. Imaging parameters must be set to yield images with a signal-to-noise ratio of at least 2 for EGFP. Start by selecting a 512 × 512 pixel image, and then set the power of the 488-nm laser to 5% of maximum (100% corresponds to 20 mW), and then choose a 60-μs dwell time on each pixel and a frame rate of 1 image per second. Adjust the laser power or dwell time parameters until a signal-to-noise (S/N) ratio of >2 is reached (with the following definition: S/N = (average of the signal)/(s.d. of the signal)). Acquire an image and, using the confocal microscope software, verify that all pixels are within the dynamic range of acquisition. Some microscope software color-codes fluorescence intensities such that pixels with values equal to zero appear in blue and pixels with values corresponding to the maximum intensity appear in red. These imaging parameters will form the starting point for determining the optimal parameters for imaging live cells.
Determination of the bleach parameters. FRAP experiments necessitate the determination of bleach pulse parameters such that photobleaching of the fluorophore is complete. This signifies that fluorescence in the ROI after the photobleaching pulse is equal to background fluorescence. The use of fixed cells ensures that diffusive recovery does not occur and hence the determination of the optimal bleach parameters is simplified. Select an imaging region that englobes the whole cell to be examined and some areas with no cell to measure background fluorescence. Acquire a first image. Calculate the background fluorescence by selecting a rectangular ROI in a region in which there is no cell. Compute the average intensity in this region using the microscope software; this is the background intensity. Next, open the photobleaching dialog box. Select a rectangular ROI of 2 ×6 μm in a plane midway through the cell height; this will be the bleaching region. Determine the initial fluorescence intensity in this ROI. Choose the 488-nm laser as the photobleaching laser, and set the photobleaching settings to maximal laser power, maximal dwell time (200 μs on the Olympus FV-1000) and 500 ms total exposure time for the photobleaching pulse duration. Note that if a 405-nm laser is also available, this can be used to bleach fluorophores, and it may prove to be more efficient in doing so. Now, manually trigger the photobleaching pulse and manually acquire a postbleach image. Determine the photobleaching efficiency by comparing the average fluorescence intensity within the ROI with the previously determined background intensity and the fluorescence intensity before photobleaching. If the average fluorescence intensity in the ROI is >15% larger than the background intensity, increase the duration of the bleach pulse and retest on a new ROI. Repeat until complete photobleaching is achieved. Conversely, if photobleaching is complete with the initial bleach settings, reduce the event time and re-test until photobleaching is no longer complete. The optimal bleaching duration is the shortest one that results in complete photobleaching within the ROI. The optimal dwell time can also be determined using a similar strategy. These photobleaching parameters need to be determined separately for each different fluorophore used (e.g., EGFP, mEmerald, EYFP and so on) and for each different microscope objective. Note that the duration of the photobleaching pulse will set the shortest characteristic time that can be detected during fluorescence recovery42.
Determination of the photobleaching volume. Analysis of experimental FRAP data in our framework requires photobleaching to take place in an isotropic region. In spatially nonisotropic photobleaching volumes, diffusion will not take place at the same rate along all spatial dimensions, and this may therefore lead to erroneous estimation of diffusion constants D (ref. 55). Although the dimensions of the ROI in the xy focal plane can be precisely set within the acquisition software, the dimension in the z direction is less well controlled and depends on the NA of the objective, the laser power and the pinhole aperture selected on the confocal microscope. Optimal photobleaching parameters need to allow both complete photobleaching of fluorescence and obtention of an isotropic photobleaching volume. Both are directly determined by the total laser power used and the duration of the photobleaching pulse. To determine the optimal parameters, we perform FRAP experiments in the cytoplasm of fixed cells. Select a rectangular ROI of 2 × 6 μm in a plane midway through the cell height, and then set the laser settings to those determined in Step 9 for photobleaching. Initially, set the pinhole to that recommended by the microscope software. Expose the ROI to one photobleaching pulse and acquire a z-stack of the imaging region (Fig. 2b). Load the z-stack into ImageJ. Measure the height of the bleached region using the 3D reconstruction and rendering plugin by clicking 'Plugins → 3d viewer'. In the 3D viewer, draw a line along the whole height of the z-stack, crossing through the ROI height using the tool from the drawing menu in the main ImageJ menu. Measure the distance between the bleached area's ridges (Fig. 2b) by clicking 'Analyze → Measure'; this is the height of the bleaching volume. Repeat this step for at least 10 cells. Ideally, an average height of 1 μm should be achieved, and this should be chosen as the diameter of the ROI in the xy plane. If the height is larger, decrease the laser power of the photobleaching pulse; if the height is smaller, increase the laser power. Next, re-determine the new bleaching height.
The smallest z extent of the point spread function achievable with high-NA and high-magnification objectives is ∼600–800 nm. Given the requirement for an isotropic ROI, this limitation defines the minimum radius of the ROI in FRAP experiments.
This calibration step must be performed at least once. It is then valid for all successive experiments using these exact photobleaching settings. As laser intensity can vary over the lifetime of the laser, it is recommended to periodically re-run Steps 8–10.
The geometry of the ROI has a great effect on the characteristic recovery time. Incorrect bleaching volume determination or anisotropic bleaching volumes will therefore lead to inaccuracies in the determination of the diffusion constants and binding constants. The assumptions underlying our analysis of experimental data are only valid for spherical ROIs.
Determination of the optimal frame rate and duration of FRAP experiments
Timing 30 min–1 h
Determining the optimal experiment duration. Acquisition of FRAP recovery curves involves the definition of the frame rate and the total duration over which recovery will be monitored. To determine the experiment duration, we perform test experiments and examine the extent and the rate of recovery. To do this, choose an imaging ROI (iROI) that contains the target organelle (for detailed criteria on how to choose the iROI, see Step 15), a photobleaching ROI centered on the organelle (Step 16) and perform a FRAP experiment (Box 1) using the settings determined in Steps 8–10. As a starting point, acquire 2–5 frames before photobleaching, and then perform the image recovery for a total duration of 200 s at a rate of 1 frame per second. Recovery of the protein examined is complete if fluorescence within the organelle of interest reaches a clear plateau before the end of the time lapse. If a plateau is reached long before the end of the time lapse, decrease the duration; if no plateau is reached, increase the duration. Repeat these measurements, adjusting the duration of the time lapse. The optimal duration is the shortest for which a clear plateau can be observed (Fig. 2d). To plot the fluorescence recovery, follow the instructions on how to load and plot data in ImageJ described in Steps 19–21. Repeat this step for 5–10 cells.
Determining the optimal frame rate to minimize imaging-induced fluorescence loss. The optimal frame rate should, in principle, be the fastest possible frame rate, but, in practice, imaging-induced loss of fluorescence during the experiment places a limitation on the frame rate for image acquisition. Too-high frame rates will decrease the final fluorescence intensity reached due to excessive fluorescence loss and affect the apparent rate of fluorescence recovery (Fig. 2d). In practice, the optimal frame rate will be such that 50–100 frames are acquired over the experiment duration and 10–20 frames are acquired during the first 50% of recovery to allow for the identification of fast recovery processes. To determine the optimal frame rate, perform two different experiments in each cell: a FRAP experiment using the conditions defined in Steps 8–10 over the duration established in Step 11 and a fluorescence-loss experiment in which an lROI will be imaged with the same conditions as those used in the FRAP experiment but with no photobleaching pulse. The lROI for the second experiment should have the same geometry as bleaching ROI for the FRAP experiment but should not be located too close to it to avoid any possible artifacts. Vary the frame rate, re-acquire a FRAP and a fluorescence-loss experiment starting with the fastest possible frame rate, and decrease it progressively until you reach a rate for which 50–100 frames are acquired over the total duration of the experiment. For each frame rate, determine the extent of imaging-induced loss of fluorescence by plotting the evolution of fluorescence intensity for a region in the center of the iROI (lROI) in the second experiment, as described in Steps 20–22. The evolution of fluorescence intensity in the lROI allows evaluation of the extent of fluorescence loss (Figs. 1b and 2d). Fluorescence decays of >40% in the lROi are acceptable. Higher levels require the user either to reduce exposure of the sample to light by reducing the frame rate, or to reduce the duration of the experiment or both. From the first experiment, determine the apparent immobile fraction for the bleaching ROI and the fluorescence in the ROI in the first frame after the photobleaching event (the initial recovery) for each different frame rate. At low frame rate, the apparent immobile fraction will be little affected by the imaging-induced fluorescence loss. With increasing frame rates, the apparent immobile fraction will increase owing to the increased imaging-induced fluorescence loss (Fig. 2d). The optimal frame rate will be a compromise between minimal imaging-induced fluorescence loss, the smallest initial recovery (Steps 13–14) and an apparent immobile fraction equal to that obtained for very low frame rates.
High frame rates may result in apparently larger immobile fractions as a result of excessive imaging-induced fluorescence loss, whereas low frame rates may result in large recovery in the first postbleaching frame and insufficient data points to characterize the recovery processes acting on short time scales (Fig. 2d).
Identifying diffusive processes of recovery. Identification of the time scale of diffusive and reactive recovery processes is crucial to allow their separation on the basis of timescales in our protocol (Fig. 2e,f). To do this, perform a new FRAP experiment on the organelle of interest with the parameters determined in Steps 8–12. Next, perform a second experiment in a ROI with a diameter that is twofold larger than the previous one. The bleach depth will also need to be adjusted, and this can be achieved by changing the pinhole opening on laser-scanning confocal microscopes or by increasing the laser power. Finally, perform a third experiment in the cytoplasm with the settings and ROI dimensions used for the first experiment. Plot the fluorescence recovery in these three experiments according to Steps 20–22. Diffusive processes should occur over fourfold larger timescales in the second experiment, as compared with the first experiment (Fig. 2e,f). Cytoplasmic diffusion is generally substantially faster than reactive recovery, but diffusion in the plane of the membrane may occur on similar or longer timescales than that in reactive processes. One strategy to minimize the contribution of diffusion in the membrane to fluorescence recovery is to choose bleach regions sufficiently large to yield recovery timescales that are far longer than the timescale of reactive recovery.
Comparing the acquisition rate with the rate of pure diffusive recovery. In general, fluorescence recovery of the protein of interest in the organelle to be studied will comprise a diffusive contribution and a reactive contribution. In the interval between the end of photobleaching and the first postbleaching frame, fluorescence recovers primarily through diffusion. Comparison of the rate of free diffusive recovery in the cytoplasm with recovery in the organelle of interest may allow separation of diffusive and reactive contributions through an appropriate choice of frame rate. To estimate the duration of free diffusive recovery, perform a FRAP experiment in the cytoplasm with the parameters determined in Steps 8–13. Plot the fluorescence recovery in the ROI according to Steps 20–22. If fluorescence does not appear to decrease much due to photobleaching in the ROI, and it does not increase much after photobleaching, the diffusion of the cytosolic pool of proteins in the ROI is many times faster than the reactive processes contributing to recovery, and we are in a situation in which diffusion and reaction are uncoupled (assuming that there is no diffusion of proteins in the bound state). If a clear recovery can be observed, then both diffusion and reaction occur on comparable timescales. In this case, we need to subtract the diffusive contribution of the cytoplasmic pool of proteins from the fluorescence recovery, and we need to perform sample recovery at high frame rates to characterize the reactive processes occurring on timescales comparable to those of diffusion. Increase the frame rate until the fluorescence loss reaches 40% or the total number of frames reaches 200 over the whole duration of the experiment using the procedure in Step 13.
Steps 11–14 must be repeated for each new protein to be examined.
Acquisition of FRAP data
Timing 5–6 h
Choosing an iROI. Choose a representative location for the iROI in the target cells. For example, the center of the cytoplasm is a good choice for a purely diffusive protein. For binding to an organelle, the iROI should have the organelle into its center to facilitate the study of the reactive recovery, and it should contain some cytoplasm so diffusive contributions can be evaluated (Figs. 1a and 2c). The iROI needs to be large enough to evaluate recovery in the bleaching ROI (typically in the center of the iROI). A separate experiment needs to be carried out in the same cell to evaluate the imaging-induced fluorescence loss in a control region sufficiently far away from the photobleaching ROI (Fig. 2c). The iROI for this experiment needs to be chosen similarly to the iROI for FRAP experiments. The diameter of the iROI will determine the maximum frame rate at which recovery can be monitored and the extent of imaging-induced photobleaching, with large iROIs leading to higher bleaching than small ones. Circular iROIs with a diameter of 2–3 μm offer a good compromise. Ensure that no autofluorescent objects or aggregates are present within the ROI (Fig. 2c). These situations can result in erroneous characteristic times and misinterpretation of the data.
Imaging the whole cell during time-lapse acquisition reduces the maximal frame rate attainable and markedly increases the extent of imaging-induced photobleaching without adding any extra information. In some cases, however, whole-cell imaging may be necessary for normalization purposes, and further calibration experiments will need to be carried out to find a suitable compromise.
Choosing a bleaching ROI. The diameter of the bleaching ROI is determined by the need for an isotropic ROI and the depth of bleaching determined in Step 10. Bleaching ROIs with diameters of 1–2 μm are optimal in our experiments.
Smaller ROIs are very sensitive to noise. The bleaching ROI should be in the middle of the iROI chosen in Step 15 and have the organelle of interest in its center (Fig. 1a). It should also contain a cytoplasmic region (the cyROI) to enable evaluation of the diffusive recovery (Fig. 2c). As for the iROI, ensure that the bleaching ROI does not contain a nonhomogeneous fluorescence distribution in the organelle to be studied or autofluorescent objects (Fig. 2c).
Identification of the mode of fluorescence recovery. After photobleaching, fluorescence in the organelle of interest can recover through diffusion, reaction or both. To identify the mode of recovery, compare the recovery curves acquired in the cytoplasm with those acquired in the organelle of interest. To do this, determine the diffusion rate of the protein by performing FRAP in the cytoplasm and in the organelle of interest for 5–10 cells. Plot the fluorescence recovery in both the conditions in the same graph as described in Steps 20–22. If, for a 1-μm bleaching ROI, the FRAP curves show a complete recovery on a timescale of 1–2 s both in the cytoplasm and in the target organelle, we can conclude that only diffusive processes are at play. Indeed, previous work has shown that cytoplasmic EGFP fluorescence recovers fully by diffusion over these timescales14. Conversely, if recovery in the cytoplasm is complete within 1–2 s but recovery in the target organelle is not, additional processes participate in recovery, and we are probably in the presence of a reaction-diffusion system.
Acquisition of a coherent set of FRAP experiments. Perform FRAP experiments using the protocol optimized in Steps 8–16. After each experiment, plot the fluorescence recovery in the ROI according to Steps 20–22 and estimate the time necessary for fluorescence to recover to 50% of its initial value, the half-time t1/2. Continue acquiring FRAP experiments until the s.d. of the half-time of fluorescence recovery changes by >10% with each additional experiment. Aim to collect data from at least 30–40 cells.
Data quality control. Fluorescence within the ROIs is sensitive to the movement of objects into and out of the ROI, as well as to movements of the cell boundaries. These movements affect the recovery curves and are independent of the turnover processes to be investigated. Therefore, each individual FRAP time lapse must be visually inspected for such artifacts after data acquisition before the data are saved and analyzed. Experiments presenting such issues should be excluded from the subsequent analysis.
Data analysis: data preparation
Timing 1–2 h
Loading of input data. Data analysis can be performed with Open-source software (Fiji/ImageJ), dedicated analysis software (Metamorph or Imaris) or software provided by microscope manufacturers. In ImageJ, open each time lapse by clicking 'File → open → file.tif' for all FRAP or loss of fluorescence curves.
Background fluorescence correction. Subtract the background fluorescence from each individual recovery and fluorescence loss curve by clicking 'Process → Subtract Background' in the main ImageJ menu. Continue the analysis with the background-corrected data. Further information on how the background subtraction plug-in functions can be found in the help menu. Note that once acquisition parameters have been set, background fluorescence signal can also be acquired from samples devoid of cells according to the procedure described in ref. 56.
Data extraction for imaging-induced loss of fluorescence curves and cytoplasmic diffusive recovery curves. Start data extraction by generating the full FRAP curves, and then repeat for the loss of fluorescence curves acquired for each cell. For the FRAP curves, position the ROI for the data analysis at the location of the photobleaching ROI. This location can be easily found by moving to the frame of the time lapse in which photobleaching took place. Draw a circular region around the photobleached region by selecting a circular region from the 'Region' menu in the main ImageJ menu. Plot the FRAP recovery curve by clicking 'Image → Stacks → Plot Z-axis Profile'. This generates a curve that gives the evolution of the average fluorescence intensity in the ROI over time. Save the data by clicking in the Results window: 'File → Save as 'Results.txt'. The corresponding graph can be saved by clicking in the graph window to select it and then clicking 'File → Save' in the Plotting window. Now repeat the same procedure to generate the imaging-induced loss of fluorescence curves. Here, choose a region containing cytoplasm in the center of the iROI. Repeat this procedure for all FRAP curves (Fig. 2d).
Data extraction for the reactive regime. Generate the fluorescence recovery curves for the whole bleaching ROI, as described in Step 22. In addition, data analysis for recovery in the organelle necessitates the generation of cytoplasmic correction curves that report on diffusive recovery within the cytoplasm. To generate these, follow the same procedure as before, but select a smaller ROI situated within the center of the cytoplasmic part of the bleaching ROI (cyROI, Fig. 2c) and plot the recovery curves according to the method described in Step 22. The full recovery curves and their corresponding cytoplasmic recoveries should be saved into the same folder as detailed in Step 21. Both display the evolution of average fluorescence intensity in their respective ROIs over time. Finally, repeat the procedure described in Step 22 to compute the general loss of fluorescence curves corresponding to each FRAP experiment. Repeat these procedures for all FRAP curves. For each FRAP experiment, this will generate a curve giving fluorescence recovery over the whole bleaching ROI, a curve giving the imaging-induced fluorescence loss and a curve giving the recovery in the cytoplasm owing to diffusion only.
Measuring the area of the cytoplasmic region within the bleaching ROI. To measure the area Acytop of the cytoplasmic region, load the time lapse into ImageJ. Choose the 'Freehand selection tool' from the 'Drawing menu' and draw a region encompassing the area occupied by cytoplasm in the ROI and click 'Analyze → Measure' to measure the area. Repeat this procedure for the total area of the bleaching ROI AROI. Perform these two measurements for all FRAP experiments and save the data in a Workbook.
Import data into analysis program. Custom-written OriginLab fitting functions are provided with this protocol (Supplementary Methods) for the analysis of the data. To load the FRAP data into Origin, use the 'ASCII import' function in the 'File' menu and choose one of the .txt files generated in Steps 20–22. Alternatively, drop all of the text files to be analyzed into the working space of the software.
Data alignment. Next, all the curves need to be aligned in time such that photobleaching occurs at time t = 0 in all curves. To do this, select the column representing time and right-click. From the menu, select 'Set Column Values'. In the dialog box, type 'Col(A)-t' with t the time at which photobleaching took place and A the selected column, and then click on 'Apply'. For example, if photobleaching occurred at time t = 2 s, we subtract 2 s from all time points of the first column by typing 'Col(A)-2' in the dialog box. This procedure must be repeated for each individual curve.
Timing 5–6 h
Depending on the regime of fluorescence recovery identified in Step 17, use option A for purely diffusive processes and option B for reaction-diffusion processes.
Purely diffusive processes • TIMING 5–6 h
Preparation. Before the analysis, copy the fitting and plotting functions provided with this protocol into the Origin program 'Fitfunc' subfolder (see Supplementary Methods and documentation for your version of Origin, as this location can differ by version).
More accurate estimates of the diffusion coefficient can be obtained from the fluorescence intensity profile in the ROI immediately after photobleaching using the methods described in ref. 9.
Data normalization. To compare recovery curves between cells, we first normalize all curves to the average intensity in the prebleach frames. To do this, select the rows corresponding to the prebleach data points, right-click and choose 'Normalize' from the menu. To plot the curves in a graph, select the desired column, right-click and choose 'Plot' from the menu.
Data averaging. The Origin software allows averaging of multiple curves using interpolation. The use of interpolation is necessary because small differences always exist between experiments in the timing of acquisition of each frame. First, drop the curves to be averaged into one Book. Select all of the curves and plot them by right-clicking and choosing 'Plot → Line' from the menu. Select the data from all curves for t > 0, and then calculate the arithmetic average, s.e.m. and s.d. by clicking 'Analysis → Mathematics → Average Multiple curves' in the main Origin menu. This will open up a dialog box. In the dialog box, click the data selection icon to the right of the Input line and select the columns to be averaged. Tick the options standard error ('Std Err'), as well as s.d. ('Std Dev'), and select an appropriate number of points for interpolation. A rule of thumb is to choose tenfold more points than frames considered.
Fitting the free diffusion data: choosing the portion of the curve to be fitted. For pure diffusion, fluorescence intensity after photobleaching recovers as a function of the form F(t) = (F(0) + F(∞) × (t/τ))/(1 + (t/τ)), where τ is the characteristic diffusion time. To fit FRAP curves, first choose the portion of the curve to be fitted (Step 27A(iv)). Next, initialize the fitting parameters (Step 27A(v)), and then perform the fitting (Step 27A(vi)). To fit the experimental data, we need to select data points ranging from the initial photobleaching event until recovery reaches a plateau (blue lines, Fig. 3a). If the recovery curve contains a long plateau, selecting all the data points within the plateau will result in placing excessive weight on the processes dominant at longer timescales. To select data points for fitting, select the window containing the graph and choose 'Data → Mark Range' from the Origin main menu. This will display two red lines that show the limits of the interval over which the data will be fitted. The interval size can be adjusted by double-clicking on the graph and dragging the red lines. Select the portion of the curve going from the photobleaching event at t = 0, when the fluorescence intensity is minimal, up to a time point when the recovery curve has reached its plateau.
For a coherent analysis, the portion of the curve fitted needs to be the same for all FRAP curves examined.
Fitting free diffusion data: initialization of the fit parameters. Select the graph window and open the nonlinear fitting dialog box by clicking 'Analysis → Fitting → Non-linear Curve Fit → Open Dialog' in the Origin main menu. Choose 'Category: User defined' and 'Function: FRAP_freediffusion'. Select the 'Parameters' tab and input initial parameters for τ, A0 and A. Figure 3a illustrates how to read these initial guesses from the FRAP curve. The amplitude A is equal to the value of the plateau at long times (i.e., 'the mobile fraction'). The offset A0 represents the initial fluorescence in the first frame after the photobleaching event and can be read off from the graph. The characteristic timescale of recovery τ can be approximated to the recovery half-time in the first instance.
Fitting free diffusion data: fitting procedure. In the dialog box, refine the initial parameter guesses by pressing the 'Parameter' button (situated under the parameter table represented by P with an arrow) up to three times. Initiate a first fit by pressing the 'fit once' button (represented by a red line crossing through scatter of blue points) and examine fit to decide whether the initial parameters are satisfactory. Unsatisfactory parameters will typically yield fits that deviate significantly from the experimental data curve in some or all parts of the graph (Fig. 3c). Next, execute fitting by pressing the 'fit until convergence' button and wait until the values for R2 and χ2 no longer improve. Good fits to the experimental data will return values of R2 ≈ 1 and χ2 ≈ 0.
Evaluate the fitting quality visually at all stages using the zoom function in the graph window (Fig. 3c).
Values of R2 > 0.5 are insufficient and indicate that experimental data need to be examined for data misalignments and inconsistencies or reacquired with a higher signal-to-noise ratio. To save the determined fitting parameters, you can either export the data as an image file, display the fitting parameters on the graph alongside the experimental curve and the fitted curve, or save the worksheet created by the fitting function that includes R2, χ2 and a plot of the residuals.
The algorithm used by the software may reach convergence but still yield a poor fit in particular at short timescales (Fig. 3c). Note that if no satisfactory fit can be obtained, this may signify that some reactive processes contribute to recovery.
Calculating the diffusion constant. Calculate the corresponding diffusion constants using the equation τdiffusion ≈ R2/D, where R is the radius of the bleaching ROI, D is the diffusion constant of the protein and τ is the characteristic time computed from the fitting in Step 27A(vi). Having now successfully fitted the average curve, repeat Step 27A(iv–vi) for all FRAP curves separately. The mean diffusion constant, the standard error and the s.d. can then be determined from the set of diffusion constants by applying standard statistical principles to the data acquired from all of the FRAP curves. Open a new data sheet from the main menu in Origin by clicking 'File → New → Worksheet' and copy all of the diffusion constants to a new column. To calculate the mean value, the standard error and the s.d., select the column by right-clicking. Next, in the main menu, choose 'Statistics → Descriptive statistics → Statistics on column' and follow the instructions provided.
Statistical relevance for data comparison. Repeat Steps 20–27A(vii) for each experimental condition to be compared. To assess the presence of statistically significant differences between the conditions, copy the diffusion constants for each condition to a separate column in a new worksheet. Next, select the columns to be compared and choose 'Statistics → Hypothesis Testing → Two-sample t-test' from the main menu. Follow the instructions for performing a t-test.
For a t-test to be valid, the data must follow a normal distribution and each sample size must be at least 30–40 data points.
Biological replicates of a same experiment should be compared statistically to ensure that they are similar and hence can be pooled.
Reaction-diffusion processes • TIMING 5–6 h
Uncoupled diffusion and reaction. If Steps 14 and 17 indicate that the FRAP experiments are in an uncoupled diffusion and reaction regime (i.e., diffusion in the cytoplasm is very fast and diffusion in the membrane is negligible, Step 13), then subtract the fluorescence intensity at time t = 0 from all data points. This will generate the reactive recovery curve. Proceed to Step 27B(iii). If Steps 14 and 17 indicate that substantial diffusive recovery occurs in the cytoplasm, then proceed to Step 27B(ii).
Diffusion correction. When diffusive recovery is not completed by the first postbleach frame, the recovery curves for the whole ROI must be corrected for cytoplasmic diffusive recovery by subtracting the diffusive recovery curves computed in Step 19 from the corresponding raw recovery curves adjusted for their relative areas. Indeed, the total fluorescence in the ROI can be expressed as Ftotal(t) = (ARO I× Fbackground(t) + (Acytoplasm) × Fcytoplasm(t) + Aorganelle × Forganelle(t))/AROI, where Ftotal is the total fluorescence intensity, AROI is the total area of the bleached region, Fbackground is the background intensity, Fcytoplasm is the cytoplasmic intensity, Acytoplasm is the area occupied only by cytoplasm, Forganelle is the fluorescence intensity in the organelle and Aorganelle is the area occupied by the organelle. Step 23 subtracts the contribution of the background from Ftotal. We now need to measure Acytoplasm and Fcytoplasm(t) to subtract the contribution of cytoplasmic recovery to the total recovery. To do this, load both the whole ROI curve and the cytoplasmic curve into Origin and add two new columns ('Right-click option → New column') to the worksheet of the FRAP curve for the whole ROI, and copy the cytoplasmic curve into the first new column. Next, subtract the cytoplasmic curve from the whole ROI curve point by point by right-clicking and selecting 'Set Column Values' from the menu. The cytoplasmic curve needs to be weighted by the relative area occupied by cytoplasm in the ROI Acytoplasm/AROI using the areas determined in Step 24. In the dialog box, type 'Col('full recovery')- Acytoplasm/AROICol('diffusive recovery')', and then click on 'Apply'. The new curve now only reports on reactive recovery processes.
The curves need to be aligned to their photobleaching events according to the procedure described in Step 26 before subtraction. Note that if the organelle of interest is bathed in cytoplasm, such as a network of F-actin14, then the fluorescence in the organelle consists of a cytoplasmic fraction and a bound fraction: Forganelle = (Fbound + Fcytoplasm), where Fbound is the fluorescence intensity of the protein bound to the organelle. By assuming a small volume exclusion for the organelle, the contribution of cytoplasmic fluorescence intensity in the organelle (Aorganelle/AROI) × Fcytoplasm must also be subtracted from Ftotal.
Data normalization. Carry out data normalization, as detailed in Step 27A(ii), on the curve generated in Steps 27B(i) or 27B(ii).
Data averaging. Carry out data averaging as detailed in Step 27A(iii). Ensure that the appropriate fitting functions are available in Origin according to Step 27A(i).
Fitting reaction data. For reactive recovery, fluorescence intensity after photobleaching recovers as the sum of i single exponential functions Fi(t) = Ai × (1−exp(−(t−t0)/τi)) with F = ΣFi.
Each amplitude Ai represents the mobile fraction of the recovery for process i. To fit reaction curves, we follow a similar protocol to Step 27A(vi). First, choose the data range to be fitted by following the instructions in Step 27A(vi).
Establish that no diffusion occurring in the plane of the membrane contributes to fluorescence recovery before proceeding with fitting. This can be done by comparing recovery in experiments acquired with different ROI diameters.
Fitting with one exponential function. Select the graph window and open the nonlinear fitting dialog box by clicking 'Analysis → Fitting → Non-linear Curve Fit → Open Dialog' in the Origin main menu. Choose 'Category: User defined' and 'Function: FRAP_onedrprocess'. Select the 'Parameters' tab and input initial parameters for ωd1, t0 and A1. Figure 3a illustrates how to read these initial guesses from the FRAP curve. The amplitude A1 is equal to the value of the plateau at long times (i.e., 'the mobile fraction'). Time point t0 should be assumed to be zero in the first instance. The characteristic timescale of recovery τ1 can be approximated to the recovery half-time in the first instance and ωd1 can be chosen as 1/τ1. Follow the instructions detailed in Step 27A(vi) to refine the fitting parameters, and perform the fit. If fitting with one process gives a fit such that R2 → 1 and χ2 → 0 that appears visually satisfactory, then save the determined parameters as explained in Step 27A(v). If the fit is not satisfactory, continue with Step 27B(vii). Figure 3c gives a visual example of an unsatisfactory fit.
Fitting with two or more exponential functions. Select the graph window and open the nonlinear fitting dialog box by clicking 'Analysis → Fitting → Non-linear Curve Fit → Open Dialog' in the Origin main menu. Choose 'Category: User defined' and 'Function: FRAP_twodrprocess'. Select the 'Parameters' tab and input initial parameters for ωd1, ωd2, t0, A1 and A2. If two processes contribute to recovery, the faster process dominates in short times, whereas the slower process dominates in longer times. To estimate these, plot the fluorescence recovery. Next, draw a box starting from t = 0 such that it reaches 30% of the total amplitude of fluorescence recovery, and a second box such that it covers the final 30% of the total recovery time (Fig. 3b). Measure the width of the first box to get an estimate of τ1 and its height to get an estimate of A1. Measure the height of the second box to get an estimate of A2 and its width to get an estimate of τ2 (Fig. 3b). Refine initial guesses as described in Step 27A(vi), and perform the fit. If this procedure gives a visually acceptable fit (Fig. 3d) such that R2 ≥ 1 and χ2≥ 0, then save the parameters as explained in Step 27A(vi). If the fit is not satisfactory, repeat the fitting with three processes using the 'FRAP_threedrprocess' for fitting (Fig. 3e). Use a similar strategy as underlined above to provide initial fit parameters for the three processes.
Increasing the number of exponential processes used will usually improve the fit quality. However, this may not necessarily signify that more processes participate in recovery. A good indication that too many processes have been considered during fitting is that at least two processes have the same reaction rates and represent a similar fraction of the total protein. The sum of all amplitudes should be equal to the sum of amplitudes from the previous fit (ΣnAI = Σn+1Aj, Fig. 3c,d). For example, fitting with one exponential function might yield A = 1, t0 = 0 and ωd = 0.1. If only one recovery process is at play, then fitting with two exponential functions will result in the following parameter set: ωd1 = ωd2, t0 = 0, A1 = 0.5 and A2 = 0.5, with the sum of A1 and A2 equal to the A found with one exponential fitting. With experimental data, the parameters sets for ωd1 and ωd2 obtained from fitting all of the data curves can be compared statistically to test whether they are significantly different from one another or not.
If the number of processes in the optimal fit is larger than the number of known binding domains, this may indicate that a diffusive process has been overlooked. If the protein of interest binds to transmembrane- or membrane-associated proteins, this may reflect a contribution from diffusion in the plane of the membrane. To ascertain this, follow the procedure detailed in Step 13, paying particular attention to longer timescales.
Reaction rates. To calculate the mean reaction rates, s.e.m. and s.d., proceed as described in Step 27A(viii).
Calculating the association rates and process fractions. If desired, association rates can be determined from the effective dissociation rates ωdi determined in Step 27B(iii) by using the equation ωa,i= ωd,i fi F/F0, where F is the fluorescence in the organelle of interest, F0 is the fluorescence in the cytoplasm and the process fraction is fi. The average fluorescence F and F0 can be determined in ImageJ using frames of the time lapse before photobleaching. To do this, follow Steps 20–22 to load the files, and extract the appropriate values. Next, compute the association rates by multiplying the dissociation rates with the corresponding process fraction fi and the ratio of the initial fluorescence intensity of the target organelle and of the cytoplasm. The process fractions fi can be calculated as fi = Ai/F, where Ai stands for the amplitudes derived from multiexponential fitting.
Data presentation: logarithmic acceleration plots
Timing 30 min
To compare the rate constants across experimental conditions and to visualize the different processes participating in recovery, we use logarithmic acceleration plots that represent the logarithm of the second derivative of the fit function (Fig. 3f). In these plots, each piecewise linear segment corresponds to a different fluorescence recovery process. The slope of each segment is ωd,i. Changes in the recovery rates in response to drug treatment or genetic perturbations will result in linear segments with steeper or shallower gradients that can easily be visualized in the logarithmic acceleration plots. To generate these plots in Origin, open a new worksheet by selecting 'File → New → Worksheet' from the main menu, and copy the fitting data including the time values into the first two columns. Right-click the fitting values and compute the second derivative by choosing 'Analysis → Mathematics → Differentiate' from the main file menu. Select '2' for order in the dialog box to compute the second derivative. Next, add an additional column and calculate the negative logarithm from the values of the second derivative by right-clicking and selecting 'Set Column Values' from the menu. In the dialog box, type 'log(-Col('second derivative'))', and then click on 'Apply'. Finally, open the logplot template provided with this protocol from the main file menu by clicking 'File → Open Template' and load the data into the graph, as previously described in Step 27A(ii).
Troubleshooting advice can be found in Table 1.
Steps 1–5, sample preparation: 6–10 h plus 48 h of culture time
Steps 6 and 7, calibration sample preparation: 30 min
Steps 8–10, calibration experiments: 5–6 h
Steps 11–14, determination of the optimal frame rate and duration of FRAP experiments in live cells: 30 min–1 h
Steps 15–19, acquisition of FRAP data: 5–6 h
Steps 20–26, data analysis: data preparation: 1–2 h
Step 27A, data analysis: purely diffusive processes: 5–6 h
Step 27B, data analysis: reaction-diffusion processes: 5–6 h
Step 28, data presentation: logarithmic acceleration plots: 30 min
Figure 4 provides an overview of the anticipated results for the application of this protocol to the ERM family protein ezrin. Ezrin-GFP localizes to the membrane-cortex interface of M2 cells (Fig. 4a). We chose an ROI centered on the membrane at the equatorial plane of the cell (Fig. 4a). We acquired 52 frame time lapses at a rate of 1 fps, with a 500-ms imaging time for each frame. Each time lapse consisted of two frames before photobleaching, and then 2 s of photobleaching was carried out using 100% laser power. Next, we followed fluorescence recovery over 50 frames. Ezrin-GFP fluorescence recovery was complete within ∼50 s (N = 25 cells, Fig. 4a,b). To determine how many molecular processes participated in ezrin recovery at the membrane-cortex interface, we fitted the experimental fluorescence recovery curves as detailed in the protocol in Step 27B(i–viii) (Fig. 4b). Three exponential functions were necessary to fit the experimental data, suggesting that three distinct molecular processes participated in ezrin turnover at the membrane-cortex interface (Fig. 4c).
To gain insight into the molecular origin of turnover processes participating in the recovery of full-length ezrin, we examined the turnover of the two principal subdomains of ezrin: the ABD and the membrane-interacting FERM domain (Fig. 1a). These constructs were generated by classic molecular cloning techniques that are not detailed in this protocol (see ref. 41), and they have been deposited in the Addgene plasmid repository (Addgene IDs: 62380, 62381, 20680). We compared the fluorescence recovery kinetics of these domains with those of full-length ezrin. Both the FERM domain and the ABD localized to the membrane-cortex interface (Fig. 4a), similar to ezrin. Therefore, we decided to acquire FRAP data with the same protocol as for ezrin. The FERM domain recovered fully over timescales comparable to that of full-length ezrin, but the ABD recovered much faster (within 2 s; Fig. 4b,c). Analysis of the recovery with multiexponential fitting showed that only two molecular processes participated in the turnover of FERM (Fig. 4c) and only one molecular process participated in turnover of ABD (Fig. 4c). The process participating in ABD recovery had a timescale similar to the first timescale of full-length ezrin. The processes leading to FERM turnover operated on timescales similar to the second and third processes participating in turnover of the full-length protein. Determination of the nature of the molecular processes leading to FERM turnover necessitated the use of single-molecule imaging techniques concentrating on qualitative aspects of speckle trajectories and speckle lifetimes, as described in ref. 15. These techniques indicated that the first turnover process of FERM resulted from association/dissociation of FERM to membrane-bound partners, whereas the second process was the result of slow diffusion within the plane of the membrane. This example illustrates the need to carefully consider diffusive processes in the membrane, as these can be misidentified as extra reaction processes.
Sprague, B.L. & McNally, J.G. FRAP analysis of binding: proper and fitting. Trends Cell Biol. 15, 84–91 (2005).
Bancaud, A., Huet, S., Rabut, G. & Ellenberg, J. Fluorescence perturbation techniques to study mobility and molecular dynamics of proteins in live cells: FRAP, photoactivation, photoconversion, and FLIP. Cold Spring Harb. Protoc. 2010 pdb.top90 (2010).
Mukhina, S., Wang, Y.L. & Murata-Hori, M. α-actinin is required for tightly regulated remodeling of the actin cortical network during cytokinesis. Dev. Cell 13, 554–565 (2007).
Wu, Y.X., Masison, D.C., Eisenberg, E. & Greene, L.E. Application of photobleaching for measuring diffusion of prion proteins in cytosol of yeast cells. Methods 39, 43–49 (2006).
Axelrod, D., Koppel, D.E., Schlessinger, J., Elson, E. & Webb, W.W. Mobility measurement by analysis of fluorescence photobleaching recovery kinetics. Biophys. J. 16, 1055–1069 (1976).
Soumpasis, D.M. Theoretical analysis of fluorescence photobleaching recovery experiments. Biophys. J. 41, 95–97 (1983).
Yguerabide, J., Schmidt, J.A. & Yguerabide, E.E. Lateral mobility in membranes as detected by fluorescence recovery after photobleaching. Biophys. J. 40, 69–75 (1982).
Ellenberg, J. et al. Nuclear membrane dynamics and reassembly in living cells: targeting of an inner nuclear membrane protein in interphase and mitosis. J. Cell Biol. 138, 1193–1206 (1997).
Kang, M., Day, C.A., Drake, K., Kenworthy, A.K. & DiBenedetto, E. A generalization of theory for two-dimensional fluorescence recovery after photobleaching applicable to confocal laser scanning microscopes. Biophys. J. 97, 1501–1511 (2009).
Goehring, N.W., Chowdhury, D., Hyman, A.A. & Grill, S.W. FRAP analysis of membrane-associated proteins: lateral diffusion and membrane-cytoplasmic exchange. Biophys. J. 99, 2443–2452 (2010).
Rabut, G., Doye, V. & Ellenberg, J. Mapping the dynamic organization of the nuclear pore complex inside single living cells. Nat. Cell Biol. 6, 1114–1121 (2004).
Al Tanoury, Z. et al. Quantitative kinetic study of the actin-bundling protein L-plastin and of its impact on actin turn-over. PLoS ONE 5, e9210 (2010).
Beaudouin, J., Mora-Bermudez, F., Klee, T., Daigle, N. & Ellenberg, J. Dissecting the contribution of diffusion and interactions to the mobility of nuclear proteins. Biophys. J. 90, 1878–1894 (2006).
Fritzsche, M., Lewalle, A., Duke, T., Kruse, K. & Charras, G. Analysis of turnover dynamics of the submembranous actin cortex. Mol. Biol. Cell 24, 757–767 (2013).
Fritzsche, M., Thorogate, R. & Charras, G. Quantitative analysis of ezrin turnover dynamics in the actin cortex. Biophys. J. 106, 343–353 (2014).
Sprague, B.L., Pego, R.L., Stavreva, D.A. & McNally, J.G. Analysis of binding reactions by fluorescence recovery after photobleaching. Biophys. J. 86, 3473–3495 (2004).
Sbalzarini, I.F., Mezzacasa, A., Helenius, A. & Koumoutsakos, P. Effects of organelle shape on fluorescence recovery after photobleaching. Biophys. J. 89, 1482–1492 (2005).
Siggia, E.D., Lippincott-Schwartz, J. & Bekiranov, S. Diffusion in inhomogeneous media: theory and simulations applied to whole cell photobleach recovery. Biophys. J. 79, 1761–1770 (2000).
Eggeling, C. et al. Direct observation of the nanoscale dynamics of membrane lipids in a living cell. Nature 457, 1159–1162 (2009).
Hofling, F. & Franosch, T. Anomalous transport in the crowded world of biological cells. Rep. Prog. Phys. 76, 046602 (2013).
Lewalle, A. et al. A phenomenological density-scaling approach to lamellipodial actin dynamics(dagger). Interface Focus 4, 20140006 (2014).
Wilson, K. et al. Mechanisms of leading edge protrusion in interstitial migration. Nat. Commun. 4, 2896 (2013).
Haustein, E. & Schwille, P. Trends in fluorescence imaging and related techniques to unravel biological information. HFSP J. 1, 169–180 (2007).
Kim, S.A., Heinze, K.G. & Schwille, P. Fluorescence correlation spectroscopy in living cells. Nat. Methods 4, 963–973 (2007).
Oh, D., Zidovska, A., Xu, Y. & Needleman, D.J. Development of time-integrated multipoint moment analysis for spatially resolved fluctuation spectroscopy with high time resolution. Biophys. J. 101, 1546–1554 (2011).
Hebert, B., Costantino, S. & Wiseman, P.W. Spatiotemporal image correlation spectroscopy (STICS) theory, verification, and application to protein velocity mapping in living CHO cells. Biophys. J. 88, 3601–3614 (2005).
Watanabe, N. Inside view of cell locomotion through single-molecule: fast F-/G-actin cycle and G-actin regulation of polymer restoration. Proc. Jpn. Acad. Ser. B Phys. Biol. Sci. 86, 62–83 (2010).
Watanabe, N. & Mitchison, T.J. Single-molecule speckle analysis of actin filament turnover in lamellipodia. Science 295, 1083–1086 (2002).
Mendoza, M.C., Besson, S. & Danuser, G. Quantitative fluorescent speckle microscopy (QFSM) to measure actin dynamics. Curr. Protoc. Cytom. 62, 2.18.1–2.18.26 (2012).
Danuser, G. & Waterman-Storer, C.M. Quantitative fluorescent speckle microscopy of cytoskeleton dynamics. Annu. Rev. Biophys. Biomol. Struct. 35, 361–387 (2006).
Mogilner, A. & Oster, G. Cell motility driven by actin polymerization. Biophys. J. 71, 3030–3045 (1996).
Holmes, W.R. & Edelstein-Keshet, L. A comparison of computational models for eukaryotic cell shape and motility. PLoS Comput. Biol. 8, e1002793 (2012).
Carlsson, A.E. & Sept, D. Mathematical modeling of cell migration. Methods Cell Biol. 84, 911–937 (2008).
Turlier, H., Audoly, B., Prost, J. & Joanny, J.F. Furrow constriction in animal cell cytokinesis. Biophys. J.l 106, 114–123 (2014).
Zumdieck, A., Kruse, K., Bringmann, H., Hyman, A.A. & Julicher, F. Stress generation and filament turnover during actin ring constriction. PLoS ONE 2, e696 (2007).
Loughlin, R., Heald, R. & Nedelec, F. A computational model predicts Xenopus meiotic spindle organization. J. Cell Biol. 191, 1239–1249 (2010).
Lenne, P.F. et al. Dynamic molecular confinement in the plasma membrane by microdomains and the cytoskeleton meshwork. EMBO J. 25, 3245–3256 (2006).
Wawrezinieck, L., Rigneault, H., Marguet, D. & Lenne, P.F. Fluorescence correlation spectroscopy diffusion laws to probe the submicron cell membrane organization. Biophys. J. 89, 4029–4042 (2005).
Smith, M.B., Kiuchi, T., Watanabe, N. & Vavylonis, D. Distributed actin turnover in the lamellipodium and FRAP kinetics. Biophys. J. 104, 247–257 (2013).
Roh-Johnson, M. et al. Triggering a cell shape change by exploiting preexisting actomyosin contractions. Science 335, 1232–1235 (2012).
Green, M.R. & Sambrook, J. Molecular Cloning 4th edn, (Cold Spring Harbor Laboratory Press, 2012).
Weiss, M. Challenges and artifacts in quantitative photobleaching experiments. Traffic 5, 662–671 (2004).
Wagner, M.L. & Tamm, L.K. Reconstituted syntaxin1a/SNAP25 interacts with negatively charged lipids as measured by lateral diffusion in planar supported bilayers. Biophys. J. 81, 266–275 (2001).
Kenworthy, A.K. Fluorescence-based methods to image palmitoylated proteins. Methods 40, 198–205 (2006).
Niv, H., Gutman, O., Kloog, Y. & Henis, Y.I. Activated K-Ras and H-Ras display different interactions with saturable nonraft sites at the surface of live cells. J. Cell Biol. 157, 865–872 (2002).
Lewalle, A. et al. A phenomenological density-scaling approach to lamellipodial actin dynamics. Interface Focus 4, 20140006 (2014).
Feder, T.J., Brust-Mascher, I., Slattery, J.P., Baird, B. & Webb, W.W. Constrained diffusion or immobile fraction on cell surfaces: a new interpretation. Biophys. J. 70, 2767–2773 (1996).
Blonk, J.C.G., Don, A., Vanaalst, H. & Birmingham, J.J. Fluorescence photobleaching recovery in the confocal scanning light-microscope. J. Microsc-Oxford 169, 363–374 (1993).
Periasamy, N. & Verkman, A.S. Analysis of fluorophore diffusion by continuous distributions of diffusion coefficients: application to photobleaching measurements of multicomponent and anomalous diffusion. Biophys. J. 75, 557–567 (1998).
Wolf, D.E. Theory of fluorescence recovery after photobleaching measurements on cylindrical surfaces. Biophys. J. 61, 487–493 (1992).
Coscoy, S. et al. Molecular analysis of microscopic ezrin dynamics by two-photon FRAP. Proc. Natl. Acad. Sci. USA 99, 12813–12818 (2002).
Tardy, Y., McGrath, J.L., Hartwig, J.H. & Dewey, C.F. Interpreting photoactivated fluorescence microscopy measurements of steady-state actin dynamics. Biophys. J. 69, 1674–1682 (1995).
Cunningham, C.C. et al. Actin-binding protein requirement for cortical stability and efficient locomotion. Science 255, 325–327 (1992).
Hao, J.J. et al. Phospholipase C-mediated hydrolysis of PIP2 releases ERM proteins from lymphocyte membrane. J. Cell Biol. 184, 451–462 (2009).
Shi, C., Cisewski, S.E., Bell, P.D. & Yao, H. Measurement of three-dimensional anisotropic diffusion by multiphoton fluorescence recovery after photobleaching. Ann. Biomed. Eng. 42, 555–565 (2014).
van Rheenen, J., Langeslag, M. & Jalink, K. Correcting confocal acquisition to optimize imaging of fluorescence resonance energy transfer by sensitized emission. Biophys. J. 86, 2517–2529 (2004).
The authors acknowledge the University College London (UCL) Comprehensive Biomedical Research Centre for generous funding of microscopy equipment. M.F. was funded by a Human Frontier of Science Program, Young investigator grant to G.C. (RGY 67/2008). G.C. was supported by a University Research Fellowship from the Royal Society.
The authors declare no competing financial interests.
About this article
Cite this article
Fritzsche, M., Charras, G. Dissecting protein reaction dynamics in living cells by fluorescence recovery after photobleaching. Nat Protoc 10, 660–680 (2015). https://doi.org/10.1038/nprot.2015.042
The cortical actin network regulates avidity-dependent binding of hyaluronan by the lymphatic vessel endothelial receptor LYVE-1
Journal of Biological Chemistry (2020)
Methods and Applications in Fluorescence (2020)
Cell Reports (2019)
Quarterly Reviews of Biophysics (2019)
Proceedings of the National Academy of Sciences (2019)