Here, we present a protocol for the simultaneous use of Förster resonance energy transfer-based tension sensors to measure protein load and fluorescence recovery after photobleaching to measure protein dynamics enabling the measurement of force-sensitive protein dynamics within living cells.
Cells sense and respond to physical cues in their environment by converting mechanical stimuli into biochemically-detectable signals in a process called mechanotransduction. A crucial step in mechanotransduction is the transmission of forces between the external and internal environments. To transmit forces, there must be a sustained, unbroken physical linkage created by a series of protein-protein interactions. For a given protein-protein interaction, mechanical load can either have no effect on the interaction, lead to faster disassociation of the interaction, or even stabilize the interaction. Understanding how molecular load dictates protein turnover in living cells can provide valuable information about the mechanical state of a protein, in turn elucidating its role in mechanotransduction. Existing techniques for measuring force-sensitive protein dynamics either lack direct measurements of protein load or rely on the measurements performed outside of the cellular context. Here, we describe a protocol for the Förster resonance energy transfer-fluorescence recovery after photobleaching (FRET-FRAP) technique, which enables the measurement of force-sensitive protein dynamics within living cells. This technique is potentially applicable to any FRET-based tension sensor, facilitating the study of force-sensitive protein dynamics in variety of subcellular structures and in different cell types.
The extracellular environment is a rich source of both biochemical and physical cues that dictate cell behavior. In particular, the physical nature of the microenvironment can mediate key cellular functions, including cell growth, migration, and differentiation1,2,3,4. Dysregulation of the mechanics of the microenvironment is a critical component to many diseases that do not yet have adequate treatments, such as cancer5, atherosclerosis6, and fibrosis7. A complete understanding of how cells convert physical stimuli into biochemically-detectable signals, a process termed mechanotransduction, requires the elucidation of the molecular mechanisms mediating force transmission, both into and out of the cells and within multiple subcellular structures.
Inside subcellular structures, proteins are constantly turning over; binding and unbinding based on the strength of their interactions with binding partners8. For forces to be successfully transmitted across a physical distance, there must be an unbroken chain of protein-protein interactions, meaning that a protein's turnover must be slow enough to sustain and transmit force to its binding partner9. While protein-protein interactions generally consist of several non-covalent bonds between the protein domains, the interaction is often conceptualized as a bound state that can transition to an unbound state under different conditions10,11. For a given protein-protein interaction, it is possible that force can have no effect on the lifetime of the interaction, known as an "ideal bond", reduce the lifetime of the interaction, known as a "slip bond", or increase the lifetime of the interaction, known as a "catch bond"10. Thus, there is an intricate relationship between protein load and protein dynamics, which we refer to as force-sensitive dynamics.
Towards understanding the effect of load on bond dynamics, a number of highly informative experiments have been performed on the single-molecule level. Using isolated proteins, or fragments of proteins and manipulation techniques such as magnetic tweezers, optical tweezers, and atomic force microscopy, these studies have demonstrated force-sensitive protein-protein interactions for several relevant proteins11,12. Both integrins13 and cadherins14, which are transmembrane proteins important for forming cell-matrix and cell-cell interactions, respectively, have demonstrated alterations in dynamics due to load. Within the cell, vinculin is recruited to both talin15 and α-catenin16 in a force-dependent manner and can form a catch bond with actin17, indicating a crucial role for vinculin at both focal adhesions (FAs) and adherens junctions (AJs) under load. Single-molecule studies allow for the isolation of specific protein-protein interactions and yield unambiguous results, but they do not account for the complexity of the cellular environment.
Landmark experiments demonstrated that several subcellular structures, including FAs and AJs, are mechanosensitive, and exhibit enhanced assembly in response to internally-generated or externally-applied loads18,19,20,21,22. Additionally, several theoretical models have suggested that mechanosensitive assembly could be driven by force-sensitive protein dynamics23,24,25. To examine these force-sensitive dynamics within living cells, a few indirect approaches have been taken. FRAP and related techniques provide a relatively simple methodology for measuring protein dynamics in cells26,27,28,29. However, the measurement of protein load has been more limited. A typical approach is to compare protein dynamics in cells with and without the exposure to a cytoskeletal inhibitor used to reduce overall cell contractility8,30,31. Conceptually, this is a comparison between a high load and low load state. However, there is no quantification of the load across the protein in either state, and there may be unintended biochemical effects of the inhibitor, such the loss of key binding sites along an F-actin filament. Another approach, specific to FAs, has been to measure total force exertion on the substrate by the FA using traction force microscopy to approximate molecular load and examine the relationship with the dynamics of a single protein within the FA32. While this approach allows for the quantification of total force, it does not provide molecularly specific information. FAs are made up of over 200 different proteins, many of which can bear load33. Thus, measuring the total force output of an FA potentially obscures the possibility of multiple force transmission pathways and does not reliably provide a measure of load on a specific protein.
Unlike previous approaches in mechanobiology, the advent of FRET-based tension sensors allows direct measurement of loads experienced by specific proteins inside living cells34,35,36. Here, we present a protocol that combines FRET-based tension sensors with FRAP-based measure of protein dynamics. We refer to this technique as FRET-FRAP. This approach enables the simultaneous measurement of protein load and protein dynamics, thus allowing the assessment of the force-sensitive protein dynamics in living cells (Figure 1). Already, the FRET-FRAP technique has been applied to the study of the force-sensitive dynamics of the mechanical linker protein vinculin37. Tension sensors have been developed for numerous proteins that are relevant in a variety of subcellular structures. For example, sensors have been developed for vinculin34 and talin38,39 in FAs, cadherins and catenins in AJs40,41,42, nesprin in the nuclear LINC complex43, α-actinin44 and filamin36 in the cytoskeleton, and MUC-1 in the glycocalyx45, among others46. Similarly, FRAP is a commonly used technique has been used on mechanosensitive proteins within the focal adhesions8,31, adherens junctions47, actin cortex26, and nucleus48. Moving forward, the FRET-FRAP technique should be broadly applicable to any of these existing sensors or newly developed sensors, allowing for the measurements of force-sensitive dynamics in a wide variety of subcellular structures and contexts. Towards this end, we provide a detailed, generalized protocol for implementing the FRET-FRAP technique applicable in these different systems. Hopefully, this will enable a wide variety of experiments elucidating the roles of various mechanosensitive proteins in regulating force transmission and in mediating cell behavior.
1. Generate Samples for Imaging
2. Set up Microscope for Imaging
3. Choose Parameters for FRET Imaging
4. Choose Parameters for FRAP Imaging
5. Acquire FRET-FRAP data
6. Analyze FRET-FRAP data
7. Interpret FRET-FRAP data
FRET-FRAP is made up of the combination of two fluorescent techniques, FRET and FRAP. As we focused on the effects of protein load, we used FRET-based tension sensors34,46. These sensors are often based on a tension sensing module consisting of two fluorescent proteins, such as mTFP1 and VenusA206K, connected by a flagelliform linker (Figure 1A). When the module is placed between the head and tail domains of a protein, it is possible to measure the load exerted across the protein. When analyzing FRET data, images taken in the acceptor channel are used to assess tension sensor localization and concentration, as this signal is independent of FRET (Figure 1B). After the calculation of FRET efficiency, protein load can be visualized on a colorimetric scale where a decrease in FRET efficiency toward cooler colors indicates an increase in protein load, and FRET efficiencies in the red range indicate low protein load (Figure 1B). FRAP imaging is conducted by using a laser to bleach the acceptor fluorophore in a single subcellular structure and monitor recovery over time (Figure 1C). The resulting normalized FRAP curve can be analyzed to extract parameters describing the protein dynamics, including the half-time of recovery and mobile fraction (Figure 1D). Because FRET and FRAP analyses were performed on the same cell, the average protein load and turnover in a subcellular structure can be plotted as a single point. Imaging multiple cells yields multiple points and an emerging trend can indicate whether a protein is destabilized (Figure 1E) or stabilized by molecular load (Figure 1F).
The vinculin tension sensor (VinTS) stably expressed in vinculin null MEFs very clearly localizes to FAs spread throughout the cell, as seen by looking at the acceptor channel image (Figure 2A). The acceptor channel image is used to create a segmentation mask that identifies each individual FA with a unique ID, visually designated by different colors (Figure 2B). The segmentation algorithm is based on the "water" method and labels the FAs approximately in order of brightness, as previously described34,65. The segmentation results are converted to a binary mask which is then applied to the FRET efficiency results (Figure 2C), and the average FRET efficiency within each unique FA is calculated (Figure 2D). Additional properties can be calculated for each FA in a similar manner, including average acceptor intensity, size, eccentricity, and location within the cell. This way, whichever FA is chosen for FRAP can be matched to the unique FA ID and the associated properties.
FRAP imaging and analysis is sensitive to several factors that can be controlled, including laser and imaging parameters, and some factors that cannot be controlled, such as overall FA stability26,27,28,29. For example, too much exposure to light during the time-lapse imaging can lead to major issues in interpreting FRAP data. Although the analysis of control FAs that were not bleached can be used to normalize for minor photobleaching over time, with too much exposure of the sample to light, the resulting FRAP curve shows an initial recovery followed by a dip in normalized intensity that cannot be accurately fit with an exponential function. If this effect is consistently observed in the data, it is necessary to re-optimize the imaging parameters to either decrease exposure time, increase the time-step between imaging frames, or decrease the length of the time-lapse to reduce the exposure of the sample to light.
Another example of FRAP data that is uninterpretable, is when the FA that was photobleached translocates rapidly during recovery28. A representative case of excessive translocation is shown in Figure 3. The initial image, where the ROIs are chosen, does not give an indication of FA stability (Figure 3A). Monitoring the bleached FA over time, it quickly moves away from the initial position and the automated tracking is unable to immediately follow due to the low fluorescent signal following photobleaching (Figure 3A). The resulting FRAP curve shows an initial phase of slight recovery with a jump when the fluorescence is recovered enough for the software to detect the FA and move the ROI (Figure 3B). This curve cannot be successfully fit by an exponential function. The rapid translocation of the FA also suggests that the FA structure is unstable. Thus, unstable FAs should not be included in the same FRET-FRAP analysis as stable FAs, due to both technical and biological issues.
With satisfactory FRET and FRAP data, the next step is completing the FRET-FRAP analysis by simultaneously assessing protein load and dynamics. Figure 4A shows the FRET efficiency maps of three vinculin null MEFs stably expressing VinTS. The FAs outlined in white were chosen for FRAP analysis, and the acceptor intensities are shown over time. These three FAs have vinculin under different amounts of load and display a different vinculin recovery profile. Quantifying these properties by calculating the half-time of recovery and plotting against the average FRET efficiency in each FA demonstrates the overall trend of vinculin being stabilized by increased load (Figure 4B). However, the mobile fraction plotted against FRET efficiency shows no trend, suggesting that mobile fraction is not regulated by molecular load (Figure 4C). Introducing a point mutation into the VinTS at amino acid 50 (A50I) has been shown to prevent vinculin binding to a major binding partner within FAs, talin66. The alteration of this protein-protein interaction affects vinculin force-sensitive dynamics. Vinculin null MEFs stably expressing VinTS A50I have different cell and FA morphologies, different vinculin loading profiles, and different vinculin dynamics (Figure 4D). Quantifying the half-times of recovery and FRET efficiencies and plotting shows that when the vinculin-talin interaction is disturbed, vinculin at FAs is destabilized by increased load (Figure 4E) while mobile fraction shows no trend (Figure 4F).
Figure 1: Principles of FRET-FRAP technique. (A) Schematic of the FRET-based tension sensor module (TSMod) inserted into a protein of interest and the effect of tension on the FRET signal. (B) To quantify FRET using sensitized emission, images are taken to capture donor signal (not shown), acceptor signal, and FRET signal. With appropriate corrections, the FRET image can be assigned a colorimetric scale to visualize how much tension is being applied to the sensor. (C) FRAP is conducted using the acceptor signal, which is directly proportional to the concentration. (D) FRAP imaging analysis produces curves of fluorescence intensity over time that can be fit using mathematical models to determine protein dynamics. (E, F) When FRET and FRAP are combined, force and turnover in a single FA can be measured. Measuring multiple FAs in multiple cells yields a relationship between protein load and protein turnover. In this analysis, a relationship in which increased load correlates with increased turnover is referred to as a force-destabilized state (E). In this analysis, a relationship in which increased load correlates with decrease turnover is referred to as a force-stabilized state (F). This figure has been modified from Rothenberg et al.37. Please click here to view a larger version of this figure.
Figure 2: FA identification and FRET analysis. (A) A vinculin null MEF expressing the VinTS visualized in the acceptor channel, where the intensity indicates local concentration of vinculin. Scale bar = 30 µm. (B) FAs are segmented based on the acceptor channel to create a FA ID mask where each FA is assigned a unique ID, here shown as different colors, approximately in order of brightness. (C) The FA ID mask is converted to a binary mask and applied to the FRET efficiency image to show the FRET efficiency values only at FAs. (D) The FRET efficiency within each FA is averaged to obtain a single value for each FA, which is associated with the FA ID in the output data table. Please click here to view a larger version of this figure.
Figure 3: Example of a translocating FA. (A) A vinculin null MEF expressing the VinTS A50I mutant sensor is visualized in the acceptor channel, with the color table inverted for clarity. Scale bar = 30 µm. The FA outlined in black was selected for bleaching. Zoomed-in images show the FA progression over time with the red outline indicating where the software identified the FA. Scale bar = 2 µm. (B) The resulting normalized FRAP curve from data in (A). There is an approximately 5% jump in intensity following point 3 resulting from the FA translocating quickly before sufficient recovery for the software to detect the change in FA location. Please click here to view a larger version of this figure.
Figure 4: Representative FRET-FRAP results. (A) Vinculin null MEFs expressing the VinTS displayed as average FRET efficiency images of the entire cell (scale bar = 30 µm) with zoomed-in inverted acceptor channel images showing FRAP recovery progression (scale bar = 2 µm). (B) FRAP half-time of recovery plotted against FRET efficiency for 32 cells, with the points representing cells in (A) highlighted in red. (C) FRAP mobile fraction plotted against FRET efficiency for the same cells in (B). (D) Vinculin null MEFs expressing the VinTS A50I mutant sensor displayed as average FRET efficiency images of the entire cell (scale bar = 30 µm) with zoomed-in inverted acceptor channel images showing FRAP recovery progression (scale bar = 2 µm). (E) FRAP half-time of recovery plotted against FRET efficiency for 21 cells, with the points representing cells in (D) highlighted in red. (F) FRAP mobile fraction plotted against FRET efficiency for the same cells in (E). Data were originally published in Rothenberg et al.37 and are visualized here in a new format. Please click here to view a larger version of this figure.
The FRET-FRAP method allows for direct measurement of force-sensitive protein dynamics, a property that has been difficult to directly probe inside living cells. The sensitivity of protein dynamics to molecular load is critical to the protein's function as a force transmitter or transducer. Loading is required for the transmission of both internally-generated and externally-applied forces, called mechanotransmission, and for the conversion of those forces into biochemically-detectable signals, called mechanotransduction. However, the alterations in load can affect the duration a protein stays bound, thus, the less time a protein spends bearing load, the less chance the force has to be transmitted to other proteins or transduced into a biochemically-detectable signal and sensed. The FRET-FRAP method bridges the gap between the molecular and cellular level by allowing molecular-scale measurements of force-sensitive dynamics to be accessed within a broader cellular context. Furthermore, it allows for these measurements to be taken while perturbing the intracellular or extracellular environment either biochemically or mechanically. This technique should be applicable to any FRET-based tension sensor, allowing for the investigation of protein mechanical state in a variety of subcellular structures and extracellular contexts.
Critical steps in ensuring that the desired FRET-FRAP measurements are obtained involve optimizing the imaging parameters and performing data analysis and interpretation. Optimizing the imaging parameters, as described within the protocol, is necessary to limit the photodamage to the sample, while allowing for the desired structures and dynamics to be distinguished and for sufficient signal strength for the calculation of FRET. Establishing these imaging parameters for a particular cell line and protein of interest early on will facilitate direct comparison between different experimental groups. It is worth noting that alterations to the system, such as mutating the protein of interest or introducing inhibitors, can lead to changes in protein localization (thereby altering signal intensity) and dynamics. The optimized parameters should enable clear, accurate measurements across all experimental conditions. Therefore, it is recommended to choose the parameters that are not at the extreme end of being useful, for example, being able to barely distinguish signal from noise.
While the imaging in this protocol was described for an epifluorescence microscope and attached FRAP laser module, FRET-FRAP is applicable to other imaging systems. For example, this technique can be adapted to line-scanning confocal microscopes as well as spinning-disk confocal microscopes with an attached photobleaching module. Imaging settings should be optimized in an analogous fashion to achieve adequate signal-to-noise without causing photodamage or excessive photobleaching. Particularly concerning FRET imaging, high quantum efficiency detectors are required to obtain sufficient signal for successful FRET calculation without inducing fluorophore damage. There are a number of publications describing separate FRET or FRAP imaging using a confocal microscope67,68,69, which can be used to guide optimization for FRET-FRAP imaging.
Following the experiment, data analysis should be treated carefully and performed in a reproducible, preferably automated, manner. Due to the inability to bleach more than 2-3 subcellular regions in a single cell before bleaching too much of the available pool of protein, the throughput of this technique is relatively limited. Thus, data sets are often combined across multiple days of imaging, requiring consistent treatment of data. Both FRET and FRAP offer challenges with data analysis. FRET index and FRET efficiency measurements allow for a quantification of protein load. FRET index is a relative measure that is highly dependent on microscope settings, while FRET efficiency measurements are absolute and independent of microscope settings55,70. We have recently shown that a previously developed method using "three-cube" imaging can be used to determine FRET efficiency from measurements of sensitized emission that are typically quantified with FRET Index when using FRET-based tension sensors56. The measurements of FRET efficiency are required if the measurements of the absolute forces experience by the tension sensors are to be calculated34. The cells expressing FRET-based tension sensors, especially stable cells at high passage numbers, may recombine or degrade the sensors, leading to unusable FRET data50. This is easily identified when calculating donor-to-acceptor ratios during the calculation of FRET efficiency37,56 but may be harder to detect using FRET index. When starting with a FRET-based tension sensor, it can be helpful to obtain a large data set (>50 cells) of only FRET data for the constructs of interest to identify the expected range of FRET efficiencies. Additionally, FRAP data may be difficult to extract from structures that are very mobile, such as FAs that are rapidly sliding or disassembling. Selecting a subpopulation of structures or optimizing cell plating conditions to mitigate this effect can help to minimize this issue.
In concept, FRET-FRAP can be applied to any FRET-based sensor in any subcellular region, with proper optimization. In practice, it may be difficult to capture force-sensitive dynamics of proteins that are not under substantial mechanical load or that have half-times of recovery on the very short timescale of a few seconds or on the long timescale of tens of minutes. Results from single-molecule studies can point to the proteins that may demonstrate force-sensitive dynamics within living cells. Thus far, this includes many FA and AJ proteins13,14,15,16 as well as some cytoskeletal elements71,72,73. Fortuitously, FRET-based sensors have been designed for many of these proteins46. These results can guide the selection of a protein of interest; however, it should not be expected that FRET-FRAP data will exactly mirror the results from these single-molecule studies. In fact, biochemical regulation, interactions with other proteins, and local cytoskeletal structure may obscure, or alter, the effects of forces on protein-protein interactions. The ability to observe these complexities is a unique strength of the FRET-FRAP approach.
A combination of manipulations to the cell and the protein of interest can be used to elucidate the important factors in regulating protein dynamics. For example, it can be helpful to have a sensor that is force-insensitive, either through the deletion or mutation of a force-binding domain35,74 as there should be no dependence of the protein turnover dynamics on the force reported by the sensor. Additionally, the mutations of other critical binding sites or phosphorylation sites in the protein can provide a more complete picture of how the protein of interest is being regulated. Making global changes to the cell or the environment through cytoskeletal inhibitors or by changing the substrate properties (ex. extracellular matrix or stiffness), respectively, can help determine how the force-sensitive dynamics of the protein respond to mechanical perturbations. Combining the information on protein load and force-sensitive dynamics with other biophysical properties of the protein can help to establish the mechanical state of the protein of interest. This can include the localization and local protein-protein interactions within a subcellular structure75,76. Additionally, the protein could reside in different conformation states, even within the same subcellular structure, depending on context76,77. Protein load, dynamics, localization, and conformation can all be simultaneously affected by internally-generated and externally-applied forces37,76,78,79, dictating a protein's role in force transmission and mechanotransduction. The versatility of the FRET-FRAP method and its potential compatibility with a variety of proteins and manipulations should enable the elucidation of the interaction between bulk mechanics, protein dynamics, and mechanosensitive signaling.
The authors have nothing to disclose.
This work was supported by a National Science Foundation CAREER Award (NSF-CMMI-14-54257) as well as grants from the American Heart Association (16GRNT30930019) and National Institutes of Health (R01GM121739-01) awarded to Dr. Brenton Hoffman and a National Science Foundation Graduate Research Fellowship awarded to Katheryn Rothenberg. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NSF or NIH.
0.05% Trypsin-EDTA | Thermo Fisher | 25300062 | |
16% Paraformaldehyde | Electron Microscopy Sciences | 30525-89-4 | |
60x Objective NA1.35 | Olympus | UPLSAPO 60XO | |
Antibiotic-Antimycotic Solution (100x) | Gibco | 15240-062 | |
Automated Stage | Prior Scientific | H117EIX3 | |
Custom Dichroic Mirror | Chroma Technology Corp | T450/514rpc | |
Custom mTFP1 Emission Filter | Chroma Technology Corp | ET485/20m | |
Custom mTFP1 Excitation Filter | Chroma Technology Corp | ET450/30x | |
Custom Venus Excitation Filter | Chroma Technology Corp | ET514/10x | |
DMEM-gfp Live Cell Visualization Medium | Sapphire | MC102 | |
Dulbecco's Modified Eagle's Medium | Sigma Aldrich | D5796 | with L-glutamine and sodium bicarbonate |
Fetal Bovine Serum | HyClone | SH30396.03 | |
Fibronectin, Human | Corning | 47743-654 | |
FRAPPA Calibration Slide | Andor | provided along with FRAPPA unit | |
FRAPPA System with 515 nm Laser | Andor | ||
Glass-bottomed Fluoro Dishes | World Precision Instruments | FD35 | |
HEK293-T Cells | ATCC | CRL-3216 | |
Hexadimethrine Bromide, Polybrene | Sigma Aldrich | H9268-5G | |
High-glucose Dulbecco's Modified Eagle's Medium | Sigma Aldrich | D6429 | |
Inverted Fluorescent Microscope | Olympus | IX83 | |
JMP Pro Software | SAS | ||
Lambda 10-3 Motorized Filter Wheels | Sutter Instruments | LB10-NW | |
LambdaLS Arc Lamp with 300W Ozone-Free Xenon Bulb | Sutter Instruments | LS/OF30 | |
Lipofectamine 2000 | Invitrogen | 11668-027 | |
MATLAB Software | Mathworks | ||
MEM Non-Essential Amino Acids | Thermo Fisher | 11140050 | |
MetaMorph for Olympus | Olympus | ||
Micro-Humidification System | Bioptechs | 130708 | |
MoFlo Astrios EQ Cell Sorter | Beckman Coulter | B25982 | |
Objective Heater Medium | Bioptechs | 150819-13 | |
OptiMEM | Thermo Fisher | 31985070 | |
Phosphate Buffered Saline | Sigma Aldrich | D8537 | |
pMD2.G Envelope Plasmid | Addgene | 12259 | |
pRRL Vector | gift from Dr. Kam Leong (Columbia University) | ||
psPax2 Packaging Plasmid | Addgene | 12260 | |
sCMOS ORCA-Flash4.0 V2 Camera | Hamamatsu Photonics | C11440-22CU | |
Sorvall Legend XT/XF Centrifuge | Thermo Fisher | 75004505 | |
Stable Z Stage Warmer | Bioptechs | 403-1926 | |
Venus Emission Filter | Semrock | FF01-571/72 |