|Home | About | Journals | Submit | Contact Us | Français|
Macromolecules drive the complex behavior of neurons. For example, channels and transporters control the movements of ions across membranes, SNAREs direct the fusion of vesicles at the synapse, and motors move cargo throughout the cell. Understanding the structure, assembly, and conformational movements of these and other neuronal proteins is essential to understanding the brain. Developments in fluorescence have allowed the architecture and dynamics of proteins to be studied in real time and in a cellular context with great accuracy. In this review, we cover classic and recent methods for studying protein structure, assembly, and dynamics with fluorescence. These methods include fluorescence and luminescence resonance energy transfer, single molecule bleaching analysis, intensity measurements, co-localization microscopy, electron transfer, and bi-molecular complementation analysis. We present the principles of these methods, highlight recent work that uses the methods, and discuss a framework for interpreting results as they apply to molecular neurobiology.
Virtually all biological processes are, at the molecular level, driven by the complex action of macromolecules. For example, in the nervous system hundreds of different proteins assemble and dynamically rearrange to produce the diverse anatomy and physiology of neurons. To fully understand these processes, it is essential to have a detailed knowledge of which molecules assemble, when, where, and how they interact, and how individual components structurally rearrange. A wealth of structural information has been acquired from a combination of biochemical, genetic, and biophysical methods. In particular, the advances in x-ray and electron crystallography and NMR have provided high resolution atomic-scale models of individual proteins and complexes. Even the structures of membrane proteins including ion channels, transporters, and receptors, which have been notoriously difficult to crystallize, are now being solved at an accelerating tempo (Gouaux and Mackinnon, 2005; Minor, 2007).
While methods such as x-ray crystallography provide atomic-scale resolution of protein structures, each structure is a snap-shot of one isolated low-energy state. Indeed, it is rare for a single protein to be crystallized in multiple biologically relevant forms and more problematic is the fact that some proteins are known to transition through rare and sometimes unstable conformational states. These movements define the specialized behaviors of proteins (Aldrich et al., 1983). Proteins also exist in the complex milieu of the cell, where local environmental factors such as lipid composition, membrane voltage, phosphorylation state, binding partners, and post-translational modifications can influence their structure. Biochemical and electrophysiological methods can provide insights into the influence of these factors on structure; however their structural resolution is coarse. Clearly, while crystallography and biochemistry can reveal a great deal about protein structure, other methods must be employed to determine the complete range of structures and movements critical to a protein's in vivo function.
Fluorescence methods have begun to fill this gap in understanding, providing a real-time view into the conformation of proteins and multi-protein complexes. Fluorescence measurements have the advantage that they can be performed in living cells and with small quantities of protein and over a wide range of time scales (picoseconds to hours). In this regard, Fluorescence provides a powerful complement to other biophysical measures of structure and when paired together these methods can reveal the complete and complex nature of a protein's structure, assembly, and dynamics.
In this primer, we review recent advances in fluorescence that permit the monitoring of the assembly and conformational movements of proteins important to neuroscience, including ion channels, transporters, motors, receptors, and SNARE proteins. We first discuss the basic physics and chemistry of fluorescent molecules and will then present fluorescent methods useful for monitoring the assembly of macromolecular complexes. Later, we discuss using fluorescence to map the intramolecular conformational movements of proteins. Throughout this review we emphasize both the strengths and weakness of fluorescence methods, providing a framework for evaluating and interpreting data generated from fluorescence methods of protein structure and function.
Fluorescent molecules absorb photons of particular wavelengths and within nanoseconds emit a longer wavelength photon (Lakowicz, 2006; Turro, 1978). Physically, an electro-magnetic field (light) can induce oscillations by resonance of the electrons in a fluorophore. The interaction between the light and the electrons can push a single paired electron from a stable ground state orbital (S0) into a higher energy excited state orbital (S1). The ability of a fluorophore to absorb a photon of a particular energy (wavelength) is related to the ease at which an electron can enter into the excited state. Experimentally this value is represented as the molar extinction coefficient (also known as the molar absorption coefficient, ε) of the molecule at a particular wavelength. In practical terms, the larger this number is, the more likely it is that a molecule will absorb the wavelength of light. While in the excited state, the electron quickly loses some energy through vibrational relaxation between high energy excited states. Thus, when this excited state electron finally falls back to the lower energy ground state, the energy released as a photon has a lower energy (i.e. longer wavelength) than the photon that was absorbed. This shift in energy between the absorbance and emission wavelength is known as the Stokes shift. These processes are rapid. Absorption of the photon occurs in 10-15 s, internal conversion happens in 10-11 s, and the lifetime of the excited state fluorophore is on the order of 10-9 s (Herman, 1998). Single fluorophores are capable of undergoing this cyclic process thousands of times before they “bleach” and can no longer emit photons. Absorbed energy is not always released as a photon. On occasion, energy is lost through heat or other processes without resulting in the release of light. The probability that an excited fluorophore will release the absorbed energy as a photon is represented by the quantum yield value (Φ). Quantum yield is a unit-less term related to the ratio of the number of photons emitted as light to the number of photons absorbed. Typical quantum yield values for common fluorophores are between 0.2 and 0.9.
The fact that fluorophores exist in the excited state for a finite amount of time makes fluorescence useful for probing the structure and conformational movements of biological molecules. Indeed, many processes that occur to excited state fluorophores including rotation of the molecule, electron transfer, fluorescence resonance energy transfer (FRET), quenching and bleaching, have been used to develop methods to examine the structure and local environment of fluorescent molecules. These methods have in turn provided a wealth of information about the structure and conformational movements of proteins to which the fluorophores are attached.
How are multimeric proteins assembled? What is the stoichiometry of protein complexes? What are the binding interactions within these complexes and how strong are they? When and where do these interactions occur inside the cell, and how fast? These are vital questions for understanding how many cellular processes work. For example, ion channels can assemble as heteromeric complexes. These complexes show unique behaviors specialized for their particular physiological role (Hille, 2001). Channels can also associate with accessory subunits. These associations can be transient and regulated, and thus modulate the behavior of the channel. At a still higher level, channels can assemble into large multi-subunit complexes containing dozens of active components. These complexes can act as organized signaling units in the cell. For example, calcium channels are thought to interact with dozens of proteins at the cell surface (Catterall and Few, 2008). Understanding the structure and dynamics of protein complexes will require methods that allow one to count subunits and probe interactions in living cells over time.
In the first section of the primer we will present classic as well as recent fluorescence methods used to test the binding and stoichiometry of protein complexes.
Most of the techniques we will discuss in this primer require labeling proteins of interest with a fluorescent dye. Many methods for labeling proteins have been developed and each has its own advantages and disadvantages. For example, cysteine chemistry can be used to attach cysteine-reactive fluorophores to sulfhydryl groups in proteins (Lundblad, 2004). Cysteine chemistry allows for targeted covalent labeling with small dyes. However, many proteins have multiple cysteines that can react with the dye. Furthermore, in cells or membrane patches many native proteins are present, each of which has modifiable cysteines. These off-target cysteines can produce large background signals (Islas and Zagotta, 2006; Zheng and Zagotta, 2000). Modifications to this method include the bi-arsenic dyes FLAsH or ReAsH which are non-fluorescent until they covalently modify four cysteines in a bent alpha helical structure (Griffin et al., 2000; Griffin et al., 1998). Because this arrangement of cysteines is rare in proteins, the problem of background is reduced.
The second general method for labeling proteins is the attachment of genetically encoded fluorescent proteins such as the different color GFP derivatives (FPs) or other intrinsically fluorescent proteins (Miyawaki, 2005; Shaner et al., 2005; Tomosugi et al., 2009; Tsutsui et al., 2008). Similar to GFP are engineered enzymes such as haloalkane dehydrogenase tags (Halo Tag) which are labeled covalently by single fluorophore-linked haloalkane groups and the 06-alkyguanine-DNA alkytransferase based tags (SNAP tag) which can catalyze the transfer of a fluorophore from a fluorophore-modified benzyl guanine to the enzyme's active site cysteine (Keppler et al., 2003; Keppler et al., 2004; Los and Wood, 2007; Maurel et al., 2008; Miller and Cornish, 2005; Reck-Peterson et al., 2006). Interestingly, a mutant of the SNAP tag (CLIP tag) has recently been developed which is specific for fluorophore-modified benzyl cytosine and can be paired with SNAP tags for two color labeling and FRET studies (Gautier et al., 2008). In addition, fluorescently-labeled antibodies can be used to bind to and mark proteins. While each of these fluorescent tags is quite specific, with almost no background fluorescence, the molecules themselves are large (e.g., GFP is ~ 40 Å by 30 Å). Other enzymes such as biotin ligase have been used to add smaller fluorescent tags directly to specific peptide sequences of proteins (Chen et al., 2005; Chen and Ting, 2005). Furthermore, techniques such as protein semi-synthesis allow the covalent addition of fluorescently-labeled peptides onto the N- or C-termini of proteins (Muir, 2003). New methods of introducing non-native fluorescent amino acids, while currently difficult, have the potential to be powerful for introducing small fluorophores into a specific location of a particular protein (Cohen et al., 2002; Pantoja et al., 2009). Finally, peptide sequences such as hexa-histidine tags or a 15-20 amino acid lanthanide binding domain can be used to non-covalently coordinate a nickel-NTA-linked fluorophore or a luminescent lanthanide ion respectively (Guignet et al., 2004; Sandtner et al., 2007).
In its simplest application, fluorescence can be used to mark the presence or location of proteins (Giepmans et al., 2006). For example, a fluorescently-labeled protein can be localized to sub-cellular compartments, such as the dendrites or axons of neurons, or intracellular compartments, such as the endoplasmic reticulum or nucleus, by means of wide-field, confocal, or total internal reflection fluorescence microscopy (TIRF). As an example, confocal and wide-field microscopy was used to monitor the subcellular distribution of a component of the store-operated calcium channel, STIM1 (Luik et al., 2008; Luik et al., 2006; Wu et al., 2006; Zhang et al., 2005). A population of fluorescently-labeled cytosolic STIM1 was seen to rapidly translocate to the plasma membrane in response to internal calcium store depletion. In another example in yeast, components of the vacuolar membrane fusion apparatus were localized to sites of membrane fusion (Wang et al., 2003; Wang et al., 2002). If two proteins interact in a complex, they both should be localized to the same region of a cell, a membrane, or a test-tube (Bolte and Cordelieres, 2006; Patel et al., 2007). To test for the association of two proteins in a cellular environment, co-localization microscopy of fluorescently-tagged proteins is a commonly used method to support a molecular interaction (Figure 1). Multicolor labeling of proteins is done generally with either paired color variants of the green fluorescent proteins (e.g. CFP/YFP or GFP/Cherry) or fluorescently-labeled antibody staining (Giepmans et al., 2006; Patterson et al., 2001). If the location of the two differently-labeled proteins overlaps, then the proteins are likely in the same subcellular neighborhood. In one beautiful example, this technique was paired with TIRF microscopy and used to map the coordinated recruitment of proteins to sites of individual clathrin coated pits during endocytosis (Figure 1)(Merrifield et al., 2002). In TIRF, a thin field of light (evanescent wave) is generated at the coverslip/water interface (Axelrod et al., 1984). This evanescent wave selectively excites fluorophores within a few hundred nanometers of the coverglass and can provide a thin optical section of the plasma membrane (Steyer and Almers, 2001). With this method several proteins present for the initiation and then completion of endocytosis were catalogued (Merrifield et al., 2002; Merrifield et al., 2005).
Recently, new super-resolution microscopy techniques have been used to map the topology of proteins with extremely high spatial resolution (Huang et al., 2009). For example, stimulated emission depletion microscopy (STED) microscopy has been used to show that the transmembrane SNARE protein syntaxin forms small clusters on the cell surface (Sieber et al., 2007). In STED, a sub-diffraction limited excitation spot is generated by two sequential laser pulses. One laser pulse that excites the fluorphores and a second longer-wavelength pulse that de-excites the fluorophores in a concentric ring surrounding the first pulse. The end result of this sequence is an effective excitation spot that is smaller than the diffraction limit of the excitation light (Hell, 2003; Klar et al., 2000). Additionally, photo-activated localization microscopy (PALM) or stochastic optical reconstruction microscopy (STORM) have been used to map the location of proteins in a cell (Betzig et al., 2006; Rust et al., 2006). In these two methods, small numbers of single fluorophores are sequentially excited and imaged. The position of each single molecule can be located to a sub-diffraction limited position (Yildiz et al., 2003). Thus, when thousands of these images are acquired and combined (reconstructed), a sub-diffraction limited image of the location of all the labeled proteins in the cell can be generated. For example, components of focal adhesions have been mapped with extremely high spatial resolution with PALM (Shroff et al., 2007; Shtengel et al., 2009).
In a more reduced system, fluorescence has been used to test for the binding of proteins and ligands to membrane proteins in isolated membranes. For example, a technique known as patch-clamp fluorometry (PCF) was used to monitor protein-ligand interactions (Biskup et al., 2007; Trudeau and Zagotta, 2004). In PCF, a membrane patch from a Xenopus oocyte expressing an ion channel is formed at the tip of a large glass micropipette (Zheng and Zagotta, 2000, 2003). The tips of these patches are then imaged with epi-fluorescence or confocal microscopy. To test for the binding of a ligand to the patch, fluorescently labeled proteins or chemicals are then perfused onto the patch. If there is an increase in fluorescence signal in the patch relative to the background that is not present in control patches, then the fluorescently-labeled molecule is likely associating with the channel. In PCF, the functional output of the channel can be simultaneously monitored with electrophysiology to correlate any binding to functional changes in the channel's behavior. This method has been used to look for the association of calmodulin with cyclic nucleotide gated ion channels, and for the binding of the activating ligand cAMP to CNGA2 channels (Biskup et al., 2007; Trudeau and Zagotta, 2004). Along with PCF, reduced membrane preparations or supported artificial bilayers containing fluorescently-labeled SNARE proteins have also been used to study the association of proteins into complexes by fluorescence co-localization analysis (Bowen et al., 2005; Brunger et al., 2009; Lang et al., 2002). Furthermore, fluorescently-labeled ATP has been used to monitor the ATP-dependent stepping behavior of myosin (Sakamoto et al., 2008).
Another technique called fluorescence correlation spectroscopy (FCS) can be used to test for the binding and association of proteins (Haustein and Schwille, 2007; Krichevsky and Bonnet, 2002). In a cuvette or on a microscope slide, a small laser spot, usually generated by a confocal microscope, is used to illuminate a dilute sample of fluorescently-tagged proteins. The average time the fluorescent molecules spend diffusing through the illuminated volume is measured from the auto-correlation of the intensity in the fluorescence signal. If a ligand binds to the test protein, the diffusion rate of the protein should be reduced. Thus, the molecule should spend slightly more time in the illuminated volume. The approach works best when the fluorescently-labeled protein is relatively small and binds to a large non-fluorescent binding-partner such as a membrane vesicle or DNA. In these cases, the mobility of the test proteins would be severely hindered upon target binding. Alternatively one can measure the diffusion coefficient of two different proteins proposed to interact. If the two molecules are interacting, their diffusion coefficients should be identical. Recent advances in FCS have great potential and can allow the real-time stoichiometry of protein complexes to be measured and mapped within cells (Digman et al., 2008; Digman et al., 2009).
Co-localization analysis in light microscopy merely assays whether two proteins are within the same general neighborhood (Piston and Kremers, 2007). Standard fluorescence microcopy cannot establish whether two proteins are directly interacting. This is because the resolution of the light microscope, dictated by the Abbe limit, is on the order of several hundred nanometers-almost 100 times the size of most proteins (~5 nanometers)(Hecht, 2002). As mentioned above, new specialized microscopic techniques such as PALM and STED microscopy are steadily improving the resolution of the light microscope (currently to ~50 nanometers), but they are still far from being able to determine whether two proteins directly interact on the molecular level. To accomplish this goal, other methods must be used. The photo-chemical process known as förster resonance energy transfer (FRET) has become a standard tool to test for the binding of two proteins (Jares-Erijman and Jovin, 2003; Piston and Kremers, 2007; Selvin, 1995; Vogel et al., 2006).
Physically, FRET is the transfer of excited state energy from one fluorophore (donor) to another (acceptor) through dipole-dipole resonance coupling (Forster, 1949). The transfer of energy is non-radiative (i.e. it does not involve a photon) and is steeply dependent on the distance between the partners. The efficiency of energy transfer is inversely related to the sixth power of the distance (R) between the pair of fluorophores (Selvin, 1995), and FRET between most commonly used fluorophores is 50% efficient at around 50 Å (the R0 value)(for a mathematical description of R0 see section 2d)(Lakowicz, 2006). In this regard, FRET is very good at determining whether two molecules are within 70 Å of one another and thus likely interacting. Over the past several years FRET has been used extensively to monitor the assembly of protein complexes. This has come about in large part due to the development of many bright monomeric color variants of green fluorescent protein (FPs)(Shaner et al., 2005). These FPs can be genetically fused to almost any protein of interest. When these fluorescently-tagged proteins are paired in appropriate combinations for FRET (CFP/YFP or GFP/Cherry for example), it can be a powerful tool to evaluate the interaction of two proteins in a cellular environment (Piston and Kremers, 2007).
While, in principle, FRET is able to precisely measure the distances between two proteins in vivo, practical limitations due to the size of fluorophores, their means of attachment, and their modes of expression, (see below) have made FRET analysis most useful for determining the dynamics of protein-protein interactions. In this regard, FRET can generally determine the affinity of a protein-protein interaction and the time course of binding and unbinding (An and Almers, 2004; Erickson et al., 2001; Erickson et al., 2003; Jensen et al., 2009; Lu et al., 2008). Importantly, however, the lack of a FRET signal does not prove that two proteins are not interacting as the fluorophores may be positioned too far away in the protein complex, quenched, or conformationally restricted from efficient participation in FRET.
To measure the interaction of two proteins with FRET, both proteins are frequently fused to one of two fluorescent probes that are good FRET partners. The most widely used FRET pair is CFP/YFP. This pair has an R0 value around 50 Å, and the fluorescent emission maxima of the donor and acceptor are relatively easy to separate with standard filter sets (Figure 2). Furthermore, new variants of CFP (mCerulean) and YFP (mVenus or mCitrine) have eliminated weak interactions between FPs and improved their photo-physical behavior to make the FPs brighter, more resistant to bleaching, and less sensitive to low pH and chloride ions (Griesbeck et al., 2001; Nagai et al., 2002; Rizzo et al., 2004; Shaner et al., 2005). These FPs are usually fused to either the N-terminus or C-terminus of the proteins of interest. Indeed, if the structure of both proteins is unknown, it is sometimes advantageous to try both orientations of the FP to rule out false-negatives (Miyawaki and Tsien, 2000; Piston and Kremers, 2007).
FRET can be measured either as an increase in the acceptor's fluorescence (sensitized emission of the acceptor) or as a decrease in the donor's fluorescence (donor quenching)(Lakowicz, 2006). One of the simplest and most widely used methods to monitor FRET in a living cell is called three-cube FRET (Figure 2)(Erickson et al., 2001; Zal and Gascoigne, 2004). Three-cube FRET can be implemented on proteins in solution, or on images from wide-field, confocal, or TIRF microscopes. The method isolates the sensitized emission of the acceptor due to FRET using a combination of three filter sets (cubes): one that measures the intensity of the donor upon excitation of the donor (IDD), one that measures the intensity of the acceptor upon excitation of the acceptor (IAA), and one that measure the intensity of the acceptor upon excitation of the donor (IDA)(Figure 2). The sensitized emission component (Fc) can be calculated by subtracting off the component due to direct excitation of the acceptor and the emission from the donor itself, using the following equation:
Where a and d are crosstalk factors that must be characterized for each instrument using acceptor-only and donor-only samples respectively.
To eliminate the dependence of Fc on the fluorophore concentration, a number of ratiometric FRET indices have been developed (Zal and Gascoigne, 2004). The two most widely used indices are FR and Eapp:
G represents the ratio of the sensitized acceptor emission (Fc) to quenched donor emission due to FRET and is a constant for a particular fluorophore pair and imaging setup (Chen et al., 2006; Koushik et al., 2006; Zal and Gascoigne, 2004). Both FR-1 and Eapp are proportional to the FRET efficiency (E) and are independent of the particular imaging setup. Therefore, the values can be compared to values acquired on other microscopes. FR-1 is also proportional to the fraction of acceptor molecules associated with donor, while Eapp is proportional to the fraction of donor molecules associated with acceptors. As such, these indices can report the degree of molecular interaction provided that the donor is in excess (FR) or the acceptor is in excess (Eapp). If these conditions are not met, little or no FRET may be observed (Zal and Gascoigne, 2004). Another commonly used acceptor sensitization method is called spectral FRET (Figure 2)(Zheng, 2006). Spectral FRET uses an index similar to FR (usually referred to as Ratio A-Ratio Ao = Fc / IAA) and is similarly proportional to the FRET efficiency and fraction of acceptor molecules associated with donor. However, with spectral FRET the entire emission spectrum of the donor and acceptor is collected and used to correct the signals and quantify the amount of FRET. Indeed, the amount of FRET (Ratio A- Ratio A0) can be calculated as a function of wavelength (Figure 2). If there is a change in this value over the spectrum then background signals or other experimental factors could be contaminating the FRET signal. To obtain spectra of a sample, experiments can be done with a fluorometer, or a microscope equipped with a spectrograph, filter array, or tunable filters (Islas and Zagotta, 2006; Taraska et al., 2009a; Taraska and Zagotta, 2007; Trudeau and Zagotta, 2004; Zheng et al., 2002).
In addition to these steady-state FRET methods, another technique commonly used to measure FRET values in living cells is Fluorescence Lifetime Imaging Microscopy (FLIM)(Lakowicz, 2006; Levitt et al., 2009). In FLIM, specialized pulsed laser excitation is used to measure the lifetime of the excited state donor fluorophore of a FRET pair. FRET introduces another pathway for loss of energy of the donor electron and thus shortens the excited state lifetime of donor fluorophores. Thus, the amount of FRET in a sample can be determined from the amount by which the lifetime of the donor is shortened in the presence of the acceptor. The advantage of FLIM is that only one fluorophore needs to be imaged in the experiment. Additionally, and more importantly, a true FRET efficiency value can be determined from FLIM experiments. For example, this technique has been applied to imaging receptor clustering into lipid microdomains and the activation of Ras and PKC enzymes (Ng et al., 1999; Varma and Mayor, 1998; Yasuda et al., 2006).
Finally, a technique known as donor de-quenching can be used to quantify FRET (Bastiaens et al., 1996). In FRET, the fluorescence of the donor is reduced (quenched) due to the transfer of energy to the acceptor. If the acceptor is photo-bleached, or destroyed, the donor can no longer transfer a portion of its excitation energy to the acceptor. Thus, by measuring the amount of fluorescence recovered in the donor (donor de-quenching) when the acceptor is photo-bleached, a direct determination of the FRET efficiency can be obtained. A drawback of donor de-quenching measurements of FRET is that the intense illumination necessary for acceptor bleaching can damage cells. However, new photo-switchable probes are currently being developed that may make multiple donor quenching/de-quenching measurements possible with time-lapse imaging (Kremers et al., 2009; Marriott et al., 2008; Patterson and Lippincott-Schwartz, 2002; Subach et al., 2009).
Several issues should be considered to ensure the correct determination of FRET. First, an extreme disparity between the levels of one fluorescent protein over the other can skew FRET measurements. Thus, careful selection of cells where the protein levels are in a reasonably similar range should be made (Piston and Kremers, 2007). Indeed, FRET is much simpler to quantify when the donor and acceptor are fused to the same molecule and are thus present in equal molar ratio (An and Almers, 2004; Miyawaki et al., 1997; Pertz et al., 2006). Second, many proteins are localized to different subcellular compartments. For example, an ion channel might be in the ER, the Golgi, and at the plasma membrane. If a binding interaction occurs only at the plasma membrane, it is important to collect fluorescence only from this region of the cell. Acquiring fluorescence from the entire cell could result in an underestimation in the amount of FRET. Furthermore, careful alignment of the imaging channels needs to be made as incorrect alignment of images can result in spurious FRET values. Depending on the magnification and optics of the system, this displacement can be larger or smaller in different regions of the image. Several analysis methods have been developed to correct for spatial differences in two-color imaging experiments (Merrifield et al., 2002; Taraska et al., 2003). This is particularly important in 3-cube FRET where a pixel-by-pixel analysis of FRET can be made.
GFP is an 11-stranded β-barrel wrapped around a central helix (Ormo et al., 1996). The chromophore of GFP forms spontaneously by the cyclization and oxidation of three amino acid residues in the middle of the buried helix (Miyawaki et al., 2003). Several techniques have taken advantage of the spontaneous formation of GFP's chromophore to study protein-protein binding interactions. One method, originally developed by Regan and colleagues, is called bimolecular fluorescence complementation (BiFC)(Figure 3)(Ghosh et al., 2000; Kerppola, 2008; Wilson et al., 2004). The technique involves splitting GFP into two fragments. These two fragments are, by themselves, not fluorescent. However, when they are brought together into a complex, the chromophore of GFP can form spontaneously to produce a bright fluorescent signal (Figure 3). At low concentrations, the two fragments cannot assemble on their own. To facilitate assembly, each piece must be fused to a partner protein that directs their assembly by bringing the two pieces together. Consequently, the amount of GFP fluorescence generated reflects the binding together of the two fusion-partners. The advantage of this method is that it is relatively simple and easy to implement and can be done in a test tube or in a cell. Furthermore, because the background signal is low, small populations of interacting proteins can be visualized even if the majority of the proteins do not interact. An interesting extension of the technique takes advantage of the different color variants of GFP to map multiple binding interactions (Hu and Kerppola, 2003; Kerppola, 2008). For example, CFP and YFP can both be split into their N- and C-terminal fragments. When the N-terminus of YFP assembles with the C-terminus of YFP, a yellow fluorescence signal is generated. Likewise, the two fragments of CFP create a cyan signal. However, if the N-terminus of YFP assembles with the C-terminus of CFP, a unique fluorophore is produced. In this regard, four interacting partners can be mapped together by monitoring which fluorescence signal is generated. Finally, work has indicated that the process of fluorophore formation can be rapid, opening the possibility that rapid associations could be mapped with this method (Demidov et al., 2006). However, dissociation rates are likely slowed by the binding energies of the two fluorescent fragments and could influence the steady-state number of interacting proteins (Hu et al., 2002). Also, the lack of a signal does not necessarily indicate that the two partners do not bind. Like FRET, if the two fragments of the fluorescent protein are positioned far away in the complex or conformational restricted from interacting, no fluorescent signal may be generated.
Many protein complexes contain more than one copy of each subunit. While FRET can hint that more than one copy of a subunit is present, it cannot easily determine the total number of each subunit in the heteromeric complex. The stoichiometry of a complex, however, can be determined from a simple method based on the relative intensities of fluorescent proteins called the Fluorescence Intensity Ratio (FIR) method (Figure 4)(Zheng and Zagotta, 2004). This method relies on the idea that when a protein, such as an ion channel subunit, is fused to a fluorescent protein both are present in equal-molar concentration. In this regard, the concentration (or corresponding intensity of light) of the attached FPs should track the concentration of the channel subunits. Thus, if there are three times more CFP –containing subunits on the cell surface than YFP-containing subunits, then there should be three times the CFP fluorescence than YFP fluorescence. In practice, the intensities of CFP and YFP are different and depend on various factors including the extinction coefficient and quantum yield of the fluorescence proteins and the intensity of the excitation light. To correct for these differences, FIR involves tagging the subunits with both possible combinations of the fluorophores. First the intensity ratio is measured from cells expressing subunit A-CFP and subunit B-YFP. Then the intensity ratio is measured from cells expressing subunit A-YFP and subunit B-CFP. The factors that affect the relative fluorescence of CFP and YFP can then be normalized out to obtain the relative ratio of subunit A and subunit B in the complex. One limitation of this approach is the FIR can only measure the relative amounts of the two subunits. It cannot distinguish, for example between stoichiometries of 3:1 or 6:2 etc. Furthermore, the experiment works best when it is done in a highly controlled protein expression system such as the Xenopus oocyte where the concentration of RNA injected results in proportional amounts of expressed protein. This technique is also not easily amenable to mammalian cells where the level of expression of two different plasmids is highly variable or the stoichiometry of the complex is heterogeneous. Furthermore, contributions to the fluorescent signals due to FRET, while generally small, can change the total amount of fluorescence from each FP and should be considered in calculations of subunit stoichiometry (Zheng and Zagotta, 2004).
All fluorescent molecules have a defined number of excitation/emission cycles before they bleach and become non-fluorescent. With the advent of single molecule techniques, single fluorophores can now be imaged routinely, even in living cells (Joo et al., 2008; Vale, 2008). When fluorophores are present at a very low density, each fluorophore will produce a well defined diffraction-limited spot (Toprak and Selvin, 2007). In single-molecule experiments, bleaching events look like rapid decreasing steps in fluorescence that do not recover. In fact, a single all-or-none bleaching event in a diffraction-limited spot is indicative of measuring a single-molecule (Pierce et al., 1997; Vale et al., 1996). If more than one fluorophore is present in a single diffraction limited spot, then multiple bleaching steps can be seen (Figure 5)(Ulbrich and Isacoff, 2007; Vale et al., 1996). In general, when many fluorophores are imaged in a single field individual bleaching events are similar in amplitude, with the average step size representing the fluorescence output from a single fluorophore.
A new technique that takes advantage of these large step-like drops in fluorescence due to bleaching is being used to answer questions of protein stoichiometry. For example, Isacoff and colleagues have used step-like bleaching events to count the number of subunits of the NMDA receptor, the voltage-activated phosphatase Ci-VSP, and the proton channel Hv1(Kohout et al., 2008; Tombola et al., 2008; Ulbrich and Isacoff, 2007). Other studies have used this technique to count the number of subunits in the Orai/STIM1 calcium channel complex and in the bacterial flagellar motor (Ji et al., 2008; Leake et al., 2006).
In practice, the protein of interest is tagged with GFP and expressed in Xenopus oocytes at very low densities on the cell surface (Ulbrich and Isacoff, 2007). When the cells are imaged with TIRF microcopy, well-defined diffraction-limited fluorescent spots can be seen. The membrane is then imaged for some time causing spots to bleach. By counting the number of bleaching steps in each fluorescent spot, the number of proteins present in a complex can be counted. Complexes that form tetramers will have up to four bleaching steps. Complexes that form dimers will produce only up to two steps. It is important to note, however, that not all GFPs fused to proteins are fluorescent. Thus, the number of bleaching steps generally follows a binomial distribution with a probability of incorporating a fluorescent GFP into each complex of less than 1. In addition, the fluorescent spots frequently need to be tracked, immobilized, or fixed to reliably follow their intensity over time (Tombola et al., 2008; Ulbrich and Isacoff, 2007). GFP can also exhibit complex blinking behaviors which can confuse the analysis of bleaching steps (Dickson et al., 1997; Garcia-Parajo et al., 2000; Pierce et al., 1997). The use of different fluorescent proteins should allow multiple components to be counted in a single complex. Furthermore, the use of photo-switchable probes will allow complexes to be tracked and bleached multiple times without destroying the chromophore completely (Marriott et al., 2008; Patterson and Lippincott-Schwartz, 2002; Subach et al., 2009).
As mentioned above, individual fluorophores generally have well defined intensity distributions. The mean of these distributions generally fits to the mean step size observed from fluorescent bleaching studies. Thus, several groups have used the raw intensity of a fluorescent signal to estimate the total number, or concentration, of molecules in a protein complex or region of the cell (Mutch et al., 2007; Sugiyama et al., 2005). First, the fluorescent signal that arises from a single fluorophore is measured. Then, the mean value of this signal is used to calibrate fluorescent signals from molecules with unknown stoichiometry. This type of imaging has been successfully done with both TIRF and epi-fluorescence microscopy. For example, this method has been used to determine the number of DAM1 subunits required to bind microtubules (Gestaut et al., 2008). An advantage of this type of imaging is that once the calibration value is determined, a single image can be used to obtain data from hundreds of molecules simultaneously without the need to perform time-lapse imaging. However, it is important to note that there can be substantial heterogeneity in the intensity of light over an area illuminated by TIRF. This can arise from several factors including the optics and illumination source used in the experiment. It is important to consider this heterogeneity in calculations of single molecule intensities. In other work, the concentration of protein in a cell was determined by quantitative immunoblot assays and then compared to the intensity of fluorescently-tagged proteins (Wu et al., 2008; Wu and Pollard, 2005). These studies were able to determine the concentration of dozens of proteins important to cell division in yeast.
Fluorescence is a powerful tool to measure the association and binding of proteins. Unique to fluorescence is the ability to monitor the assembly and binding of small numbers of complexes in a cellular environment over time. In this regard, the dynamic, transient, and regulated association of proteins can be explored. As new imaging techniques, dyes, and analysis are developed these methods will only increase in their power.
Proteins are by nature flexible and their structures make excursions (Henzler-Wildman and Kern, 2007). Proteins move between numerous low energy conformational states and these distinct structural states are what generate their intricate and specialized functional behavior. Indeed, all proteins that are regulated are controlled in some way by changes to their underlying structure. Understanding these changes is a major goal of modern biochemistry. Fluorescence techniques can help reveal the structure of proteins, and how these structures change. In the second section of this primer we review methods for studying the structure and molecular movements of proteins. In particular, we discuss how techniques such as quenching, FRET, LRET, polarization, and electron transfer have been harnessed to measure the structure of individual residues and domains within proteins.
The core of most proteins is non-polar while the outside shell of residues is exposed to the polar solvent (Creighton, 1993). For integral membrane proteins or membrane-associated proteins, a portion of the protein's structure is usually embedded in the hydrophobic lipid bilayer. The quantum yield and spectral characteristics of some fluorophores is very sensitive to the polarity of the environment. For example, the emission spectra of dyes such as coumarin, NBD, AEDANS, and tryptophan become significantly blue-shifted and their quantum yields increase in non-polar environments compared to polar ones (Lakowicz, 2006). This environmental sensitivity has been utilized to characterize the local environment, as well as changes to the local environment, of proteins (Loving et al., 2009).
One of the most useful environmentally-sensitive fluorophores is the native amino acid tryptophan (Teale, 1960; Teale and Weber, 1957). Tryptophan's emission maximum can shift between 335 nm in non-polar environments to 355 nm in polar environments. These shifts, along with changes to tryptophan's intensity, have been used to study the folding and conformational changes of proteins containing native or engineered tryptophans. For example, tryptophan has been used to study the SNARE-binding protein synaptotagmin (Chapman and Davis, 1998). In this work, tryptophan residues were introduced into the cytosolic C2 domain of synaptotagmin. Changes in fluorescence were then used to follow the calcium-dependent association of the domain with the lipid bilayer. When the domain containing the tryptophan associated with the membrane, a substantial blue shift and increase in fluorescence intensity was observed. These studies demonstrated that particular loops within the C2A domain of synaptotagmin dive into the membrane in response to calcium-binding. Similar experiments have been done with the cysteine-reactive environmentally-sensitive fluorophore AEDANS (Bai et al., 2002). Likewise, the fluorophore NBD has been used to test for the association of dynamin's pleckstrin homology domain with membranes (Ramachandran et al., 2009). NBD fluorescence has also been used to probe the membrane structure of the mitochondrial inner membrane protein translocation complex Tim23 (Alder et al., 2008).
Environmentally-induced changes in fluorescence have also been used in ion channels to study conformational changes. For example, in voltage-gated potassium channels, movements of the channel have been studied by following the fluorescence of dye-modified cysteines residues as the channel was opened and closed (Figure 6)(Mannuzzu et al., 1996). By following the voltage-dependent changes in the dyes intensity, due to changes in the fluorophore's quantum yield, these studies were able to map which portions of the channel rearranged during channel activation (Cha and Bezanilla, 1997; Cha et al., 1999; Mannuzzu et al., 1996; Pathak et al., 2007). It is often difficult to glean much information from the observation that one fluorophore-labeled residue exhibits a large change in fluorescence. However, if multiple residues localized to a specific region show similar changes in fluorescence then it is likely that the structural domain is rearranging during the conformational transition (Figure 6). This was seen for potassium channels, where dye-modified residues with large changes in fluorescence tended to cluster in specific locations of the protein (Figure 6)(Pathak et al., 2007). Similar studies have been done with the voltage-activated phosphatase Ci-VSP and other channels and transporters (Geibel et al., 2003; Gonzalez et al., 2009; Kohout et al., 2008; Larsson et al., 2004; Pless and Lynch, 2009; Tombola et al., 2009; Villalba-Galea et al., 2009; Villalba-Galea et al., 2008).
A combination of single-molecule fluorescence microscopy and environmentally sensitive fluorophores has been used to look at conformational changes in even single voltage-gated potassium channels in living cells. In these studies, TIRF microscopy was used to image single dye-labeled ion channels on the surface of Xenopus oocytes (Sonnleitner et al., 2002). Channel fluorescence underwent large voltage-dependent changes in fluorescence during channel activation. These studies demonstrated that the behavior of even single membrane proteins can be monitored over time in a cellular environment with fluorescence.
Aside from environmentally-induced changes in fluorescence, many extrinsic chemicals are able to quench the emission of a fluorophore (Lakowicz, 2006). Quenching can occur in two ways classified as either static or dynamic. In static quenching, the quencher makes a stable ground-state complex with the fluorophore and prevents the fluorophore from entering the excited state. In dynamic quenching, the quencher transiently contacts the excited state fluorophore and provides a route for the excited state fluorophore to lose energy without emitting a photon. In either case, the end result is that less photons are emitted from the fluorophore in the presence of the quencher. Both processes require intimate contact between the quencher and the fluorophore. Because close contact with the quencher is required, dynamic quenching with water soluble quenchers has been utilized to look for the solvent exposure of fluorophore-labeled residues (Lakowicz, 2006).
With fluorescent quenching experiments, the accessibility of the fluorophore is determined by measuring the fluorescence of the dye in the presence of increasing concentrations of quencher. The fluorophore's fluorescence is then plotted against 1/[quencher] to create a Stern-Volmer plot (Lakowicz, 2006; Lehrer, 1971). This type of plot can then used to test for the degree of exposure of the residue. Changes in quenching that occur in different states of the protein (with or without activating ligand for example) indicate changes in the exposure of the residue to solvent during the conformational change. In some studies, the use of differently-charged quenchers such as iodide (negative) and thallium (positive) has been employed to look for the proximity of charged groups to labeled-residues (Zheng and Zagotta, 2000).
Occasionally, quenching can occur when two of the same fluorophores interact with each other. For example changes in rhodamine fluorescence due to dye-dye quenching has been used to look for conformational changes in proteins (Grama et al., 2001). Also the dye bodipy shows significant distance-dependent self quenching has been used to test for the assembly of dynamin into higher order oligomers (Ramachandran et al., 2007). A related phenomena known as excimer fluorescence has been used to test for the proximity between two dyes. In excimer fluorescence, the stacking of the same dye changes the emission spectrum of the probes (Betcher-Lange and Lehrer, 1978; Graceffa and Lehrer, 1980; Lehrer, 1997). For example, the dye pyrene shows a dramatic red shift in its emission spectrum when two pyrene rings stack and has been used to monitor the polymerization of actin filaments (Kouyama and Mihashi, 1981).
Many domains in proteins are dynamic (Henzler-Wildman and Kern, 2007). These fluctuations in protein structures can be detected by measuring the orientation dependence of fluorescence (i.e. fluorescence anisotropy). Specifically, if a solution of fluorophores is excited with polarized light, a population of molecules is preferentially excited. These preferentially excited dyes are the fluorophores whose absorbance dipoles have a component that aligns with the polarization of excitation. Fluorophores also have geometrically distinct emission dipoles. Thus, the emitted light generated from the excited population is also strongly polarized. However, fluorophores generally re-orient in space during the excited state and because of this re-orientation, the angle at which photons are emitted becomes randomized. The more the molecule moves, the more this randomization or de-polarization occurs and the lower the anisotropy.
In practice, anisotropy is measured by first exciting the sample with polarized light. The intensity of the emission light is then measured through two polarizers, one parallel (I) and one perpendicular (I) to the orientation of excitation. From these measurements, the anisotropy can be calculated from the equation: . G is a correction factor specific to each experimental setup. Anisotropy values range from zero (high mobility during the excited state lifetime of the fluorophore) to 0.4 (essentially immobile). The time range of motion that can be examined with this method is limited to the lifetime of the particular dye. For example, most fluorophores have lifetimes from 1 to 10 ns. Some phosphorescent probes have very long excited state lifetimes (> 1msec) and have been used to look at motions that are slow on the molecular scale. For example, the motion of the myosin head group during stepping was examined with anisotropy decay measurements of the phosphorescent probe erythrosine (Ramachandran and Thomas, 1999). As an extention, fluorescence-depolarization anisotropy measurements allows the anisotropy from multiple populations of fluorophores in the same sample to be measured (Hibbs et al., 2006; Lakowicz et al., 1980; Munro et al., 1979; Thaler et al., 2009).
Polarization has also been used to study the sub-cellular structural rearrangements of proteins in living cells. For example, polarization has been used to study the rearrangements of septin proteins involved in cytokinesis (Vrabioiu and Mitchison, 2006). Another exciting development is the combination of polarization measurements with total internal reflection fluorescence microscopy (TIRF) and even single molecule measurements (Axelrod et al., 1984; Sund et al., 1999). If the light used to generate an evanescent field is polarized, then the resulting field is also polarized (Sund et al., 1999). By monitoring the brightness of fluorophores in relation to the polarization of excitation, it is possible to determine the angle of the absorption dipole relative to the cover slip. This technique has been used to monitor the orientation of lipids in membranes and dyes cross-linked to single myosin motors or actin filaments (Beausang et al., 2008; Forkey et al., 2003; Quinlan et al., 2005; Sund et al., 1999; Taraska and Almers, 2004; Zenisek et al., 2002). It is essential for both of these methods that the rotational mobility of the fluorophores attached to the protein should be limited. For example, with GFP, a continuous alpha helix was used to link GFP's N-terminus to the septin protein (Vrabioiu and Mitchison, 2006). And with myosin, a bi-functional rhodamine dye was used to lock the fluorophore down to two cysteine residues within the protein (Forkey et al., 2003).
The above methods, while capable of producing useful structural information cannot map distances. One of the most powerful tools for examining the structural rearrangements of a protein is FRET (Stryer and Haugland, 1967). In the first section of this primer we discussed using FRET to measure the binding and association of protein complexes. In this next section, we review FRET techniques for specifically monitoring distances and distances changes within proteins and protein domains.
To obtain distance measurements from FRET several values must be obtained. First the intrinsic R0 value of the pair of dyes (distance at which FRET is 50% efficient) must be known. This number can be calculated from experimentally observable constants with the equation:
where J is the normalized spectral overlap between the donor's emission and acceptor's absorbance, n is the refractive index of the solvent, qd is the quantum yield of the donor, and κ2 is a geometric factor related to the relative angle of the two transition dipoles (Selvin, 1995). Second, the FRET efficiency (E) of the sample must be determined. This value is essentially the amount of fluorescence lost in the donor when it is in the presence of the acceptor:
FDA is the fluorescence of the donor in the presence of the acceptor and FD is the fluorescence of the donor alone. With these two values (E and R0) the distance (R) between the pair of the dyes in the experiment can be calculated from the equation:
The first consideration in using FRET to map a structure is the specificity of labeling. Fluorophores must be introduced into unique locations of the protein with high efficiency while at the same time minimizing background labeling. The use of genetically fused fluorescent proteins has made it relatively easy to specifically label a protein. However, green fluorescent protein or other protein based fluorophores such as Halo tag are large (~30 Å). This is comparable to the size of small proteins themselves. Furthermore, introducing a bulky GFP into a protein structure can change the behavior and gross overall structure of the protein (Giraldez et al., 2005). Thus, protein-based fluorophores alone are not optimal labels to study the specific structural rearrangements of domains within proteins.
The most common way to add small molecule fluorophores to a protein is through cysteine modification (Lundblad, 2004). However, currently it is difficult to specifically label a single cysteine in the background of a cell where there are thousands of native modifiable cysteines. These native cyteines can produce enormous background signals. Indeed, if the protein of interest is present in relatively low numbers, background signals can overwhelm signals emanating from the protein of interest. One method to reduce this background fluorescence is to block background cysteines with a non-fluorescent cysteine reactive compound before expressing and modifying the protein of interest (Mannuzzu et al., 1996). This is a particularly powerful approach when the cysteine of interest exhibits state-specific modification. Under these conditions, the background cysteines can be blocked in the non-permissive state, and then the fluorophore can be reacted specifically in the permissive state (Islas and Zagotta, 2006; Zheng and Zagotta, 2000).
Another approach to achieve specificity is to measure FRET between a GFP donor and a small molecule fluorescent acceptor (Figure 7). Under these conditions the specific fluorescence from the donor can be used to measure FRET even if there is a large and mostly non-specific background acceptor signal. This combination allows the measurement of distances between small molecule acceptors attached at specific sites in the protein relative to the donor GFP tag. Depending on the cysteine-reactive fluorophore used, the GFP and protein of interest may need to have their endogenous reactive cysteines removed (Ostergaard et al., 2001; Taraska and Zagotta, 2007). Futhermore, the GFP is generally attached to either the N- or C-terminus of the protein to prevent disruption of the protein's overall structure and activity.
This approach was used to image the structure and conformational movements of a cyclic nucleotide-gated ion channel CNGA1(Figure 7)(Taraska and Zagotta, 2007). In this work, individual cysteines were introduced into an otherwise cysteine-less channel and labeled with the cysteine-reactive fluorophore Alexa-568 in excised inside-out membrane patches (Matulef et al., 1999). A GFP tag that could not react with cysteine-reactive dyes was fused to the C-terminus of the channel and served as the FRET donor. The absolute FRET efficiency between each dye-modified cysteine and the GFP was determined by monitoring the amount of GFP fluorescence that was lost (quenched) when the acceptor dye was reacted with a single cysteine in the channel (Figure 7). The distance each of these dye-modified cysteines moved during channel activation was then determined from changes in the FRET signals upon application of activating ligand. Using PCF, the changes in structure measured with FRET were recorded simultaneously with the changes in function measured with ionic current. From this work, the structural rearrangements in the cytosolic gating ring resulting from ligand-binding were mapped in intact membrane-resident channels (Taraska and Zagotta, 2007).
GFP and its derivatives have been used as donors with other small acceptor dyes. For example, in G-protein coupled receptors (GPCR), a CFP donor was attached to the C-terminus of the protein and used to study the structure of domains labeled with the green tetra-cysteine-reactive fluorophore FLAsH (Griffin et al., 2000; Hoffmann et al., 2005). To measure FRET, the quenching of donor CFP upon addition of the FLASH tag was used to determine FRET efficiencies. These measurements are conceptually similar to donor de-quenching measurements except that the acceptor does not have to be bleached. Large changes in the steady-state FRET signal were detected when the receptor was activated, demonstrating a structural rearrangement in this region of the protein during receptor activation.
FRET between two cysteine reactive organic dyes has been used to monitor the structure of proteins (Brunger et al., 2009; Corry et al., 2005; Ferreon et al., 2009; Glauner et al., 1999; Koch and Larsson, 2005; Larsson et al., 2004; Margittai et al., 2003; Tomishige et al., 2006). In this arrangement, methods for controlling the amount of donor and acceptor added to either cysteine are essential to produce meaningful data. This is generally accomplished by adding one of the probes (either the donor or acceptor) in much lower concentrations than the other. Thus the chance that two of the same probes are added to one protein is greatly reduced. For acceptor sensitization experiments, the limiting probe should be the acceptor, while for donor de-quenching experiments, the limiting probe should be the donor. However, other systems have been used to control the ratio of labeling including exploiting the fact that some cysteines have disparate reaction kinetics. Thus, the cysteines that react faster can be labeled first with an acceptor and the donor can be added later to modify the remaining slower-reacting cysteines (Koch and Larsson, 2005). In another example, in purified GPCRs, FRET between a di-cysteine-reactive FLAsH and the single cysteine-reactive dye Alexa-568 was used to monitor conformational changes in the receptor during ligand activation (Granier et al., 2007). Other schemes involving single molecule analysis have been used to separate molecules that have two donors, a donor and acceptor, or two acceptors respectively (Ferreon et al., 2009; Margittai et al., 2003). This technique has been particularly useful for studying the assembly and structural transitions of the synaptic SNARE complex in supported lipid bilayers (Bowen et al., 2005; Brunger et al., 2009; Weninger et al., 2008; Weninger et al., 2003). For all these methods it is essential that the ratio of donor to acceptor be known to accurately transform FRET efficiencies into distance measurements.
Another FRET technique for monitoring the structure and movements of membrane proteins takes advantage of the intrinsic orientation of membrane proteins relative to the plane of the plasma membrane. For example, membrane resident dyes and non-fluorescent quenchers can be used as FRET acceptors to fluorophore-labeled membrane proteins (Chanda et al., 2005a; Chanda et al., 2005b). In this method, the membrane-resident dyes act as a fixed point of reference. For example, in potassium channels, FRET between dye modified channels and the membrane probe dipicrylamine (DPA) was used to look at conformational changes in the voltage sensor domain during channel activation (Chanda et al., 2005a). In cyclic nucleotide-gated ion channels the movements of the cytosolic gating ring were studied with FRET between protein-linked fluorophores and membrane bound DPA (Taraska and Zagotta, 2007). Futhermore, FRET between membrane-localized dyes and fluorophore-labeled integrins has also been used to watch conformational changes in these large extracellular proteins (Chigaev et al., 2003).
A closely related technique to FRET is lanthanide or luminescent resonance energy transfer (LRET)(Selvin, 2002). Like FRET, LRET involves the distance-dependent transfer of energy from an excited state donor to an acceptor. The process has the same inverse sixth-power of distance dependence as FRET and similar equations can be used to transform LRET efficiencies into distances. As a result, like FRET, LRET has been used to monitor the structure and conformational movements of proteins. LRET differs from FRET in that a luminescent lanthanide metal ion is used as the donor instead of an organic fluorophore or FP (Figure 8). These ions have small extinction coefficients and thus generally need to be chelated by an absorptive “antenna” chromophore (Selvin and Hearst, 1994). These antenna groups absorb excitation light and transfer the energy to the lanthanide ion, boosting the lanthanide's luminescence. Commonly, the chelates contain a reactive group that allows the metal-ligand complex to be attached to an amino acid side chain. Aside from these differences, experimental data generated from LRET measurements is comparable to data generated from FRET and distance measurements can be obtained from both techniques. In most cases, organic fluorophores such as rhodamine or fluorescein are used as the acceptor for LRET studies.
LRET is an attractive method for studying protein structure for several reasons. First, the efficiency of energy transfer is dependent on the relative orientation of the transition dipoles of the donor and acceptor (Dale et al., 1979). Because lanthanide ions have multiple emission dipoles the chance that the orientation of the probe will have a large impact on the efficiency of energy transfer is less than that of an organic fluorophore (Selvin, 2002). Thus, it is safer to assume the 2/3 κ2 value generally applied to the Förster equation. A second advantage of LRET is that some probe combinations have extremely long R0 values. These R0 values can approach 100 Å. Thus, if it is necessary to position the two probes very far apart, LRET is a good option for these structural measurements. The third advantage of LRET is that the extremely long excited state lifetime of lanthanide probes allows for more straightforward measurements of donor or acceptor lifetimes. Measurements of lifetimes are more resistant to background fluorescence and can enable detection of multiple populations of protein structures within the same experiment. However, the extremely long lifetimes of lanthanides can also adversely affect measurements of distance (see below)(Tombola et al., 2006).
LRET has been used to characterize the structural transition in many proteins. For example, LRET has been used to monitor the conformational movements of voltage-gated potassium channels in oocyte membranes (Posson et al., 2005; Posson and Selvin, 2008). In these experiments individual cysteines were introduced into extracellular domains of the channel. These residues were labeled with terbium chelates, and the acceptors were subseqently added to the channels by introducing a rhodamine, fluorescein, or bodipy-labeled scorpion toxin that binds to the channel's extracellular pore (Figure 8). Distances were then mapped between cysteines labeled with lanthanide chelates and the pore-bound fluorescent toxin. The channels were then activated by shifting the voltage to depolarizing potentials and the lifetime of the lanthanide luminescence was measured. In this configuration, if energy transfer is occurring, then the fluorescence lifetime of the donor should decrease as the FRET efficiency increases. The lifetime measurements can then be used to calculate the absolute energy transfer efficiency and the position and change in position of the lanthanide during channel activation. From these studies, distance changes in the voltage sensor were characterized allowing the conformational route the voltage sensor travels during channel activation to be mapped. LRET has also been used to measure the conformational movements of myosin (Burmeister Getz et al., 1998; Xiao et al., 1998).
There are several important aspects of both FRET and LRET that have limited their use to accurately track conformational changes in proteins. First, the R0 values for most commonly-used resonance energy transfer probes are quite long (30-80 Å) compared to the size of many proteins and protein domains. While having a large R0 is advantageous for binding experiments where one might want to reduce the chance of false negatives, it can be troublesome for studies of intra-molecular movement of proteins. Indeed, with a large R0 the probes must be spaced at long molecular distances to be in the sensitive range for FRET. A second problem with traditional FRET and LRET measurements is that commonly used probes such as Cy3 and Cy5, Alexa dyes, terbium chelates, and fluorescent proteins are quite large. They are also generally attached to proteins by long, flexible, multi-carbon linkers. These linkers can distribute the position of the dyes in space and significantly skew FRET measurements of absolute distance (Lakowicz, 2006; Lakowicz et al., 1990). As mentioned, the orientation dependence of FRET can also be a complicating issue in generating accurate FRET measurements of distance (Dale et al., 1979). It is also sometimes difficult to differentially label two sites within one protein for either FRET or LRET measurements. Finally, for LRET in particular, the long excited state lifetimes of the luminescent lanthanides allows for substantial conformational movements in the protein and dye. Thus, LRET can potentially bias measurements towards distances of closet approach instead of the average position of the probes in space (Chakrabarty et al., 2002; Posson and Selvin, 2008; Tombola et al., 2006).
A technique called transition metal ion FRET addresses several of the complicating issues associated with traditional resonance energy transfer methods. In transition metal ion FRET, the donor is generally a small cysteine-reactive organic fluorophore such as fluorescein or bimane (Figure 9)(Taraska et al., 2009a; Taraska et al., 2009b). Monobromobimane in particular is extremely small, comparable to the size of a natural amino acid (Kosower et al., 1979). Additionally, these probes have short linkers that retain the conformational flexibility of the dye. Since both probe size and linker lengths are kept to a minimum, the dye is positioned close to the peptide backbone and the location of the dye more accurately reflects the location of the protein backbone.
In transition metal ion FRET, the acceptor is a transition metal ion such as Ni2+, Cu2+, or Co2+ (Figure 9). These ions absorb visible wavelengths of light and can act as energy acceptors (Horrocks et al., 1975; Richmond et al., 2000; Sandtner et al., 2007; Taraska et al., 2009a; Taraska et al., 2009b). The metal ion acceptors are also attached close to the protein backbone. This can be accomplished in at least two ways. First, two histidine residues can be engineered into an α-helix or β-sheet via mutagenesis. Di-histidine motifs can bind transition metal ions with high affinity (Arnold and Haymore, 1991; Suh et al., 1991). The second method of attaching the metal acceptor to the protein is by introducing cysteine-reactive metal binding groups such as NTA or EDTA (Dvoretsky et al., 2002; Taraska et al., 2009b). These organic chelators are capable of binding transition metals with extremely high affinities. However, these chemical groups tend to be longer and more flexible than di-histidine groups (Taraska et al., 2009b).
In transition metal ion FRET, like LRET, the metal ion probes have multiple absorption dipoles (Figgis and Hitchman, 2000). This decreases the orientation-dependence of energy transfer measurements (Horrocks et al., 1975). Metal ion acceptors (Ni2+, Cu2+, or Co2+) also have unique absorbances in the visible range. Thus, each metal ion can bind to di-histidine motifs or metal chelators and produces a unique distance dependence. As a result, different acceptors can be used in the same experiment to tune the FRET pair for measurements of distance. The R0 values with transition metal ion FRET are short, ranging between 9 and 16 Å for most dye-metal pairs. This allows close-range distances to be mapped within protein domains. A large subset of molecular movements consists of subtle changes in structure associated with the limited displacement of domains and secondary structural elements. Transition metal ion FRET is well suited to detect these short-range movements.
Recently, metal ions have been used in combination with lanthanide donors in LRET experiments to map short-range conformational movements of potassium channels (Sandtner et al., 2007). Transition metal ions were bound to sequences of six histidines engineered into extracellular loops of the channel. Transition metal ion FRET has also been used to track conformational movements of the gating ring of HCN2, a cyclic nucleotide modulated ion channel (Taraska et al., 2009a). In these experiments, individual cysteines were introduced into an otherwise soluble cysteine-less C-terminal domain of the channel and labeled with the cysteine reactive dye fluorescein-5-maleimide. Pairs of histidines were then engineered into a helical domain proposed to cap the cyclic nucleotide binding site. With FRET measurements of distance, the structure and conformational movements of the HCN2 C-terminus were mapped during ligand-binding (Figure 9).
A unique aspect of transition metal ion FRET is that it is capable of monitoring a protein's secondary structure (Taraska et al., 2009a). This is possible because the affinities of metal binding sites used in transition metal ion FRET are sensitive to helix stability (Arnold and Haymore, 1991; Kellis et al., 1991; Suh et al., 1991; Todd et al., 1991). Thus, changes in helix stability will be reflected in changes in metal affinity. In this regard, transition metal ion FRET permits two different but complementary types of information to be gathered in a single experiment. First, the amount of FRET at saturating metal concentrations provides information about the distance between the probes and second, the affinity of the site for metal provides information about the stability or structure of the metal binding site. If there is a change in the organization or stability of the helix, then this change will be seen in measurements of metal ion affinity. Indeed, in HCN2, the binding of cAMP cause an increase in the affinity at all the metal ion binding sites in the capping C-helix (Figure 9)(Taraska et al., 2009a). These data demonstrated that the binding of ligand enhanced the helicity of the C-helix.
Like other fluorescence methods there are several import aspects of transition metal ion FRET that should be considered when it is applied to structural studies. First, using the technique inside a cell is not yet possible as adding or removing metal ions into a cell is troublesome. Thus, the method is currently limited to proteins in solution, isolated membrane preparations, or to extracellular domains of proteins. Second, some proteins have endogenous metal ion binding sites and the binding of transition metal ions to these sites can influence the behavior of the protein (Waldron et al., 2009). This is true for engineered sites as well (Johnson and Zagotta, 2001; Lu et al., 2009; Webster et al., 2004). Native sites can also contribute to FRET if they are located close to the introduced fluorophore and their contribution to FRET should be accounted for in measurements of distance (Taraska et al., 2009a). Finally, metal ion binding sites have unique affinities so that full dose response curves should be performed to ensure that the metal sites are fully occupied (Arnold and Haymore, 1991; Taraska et al., 2009a; Taraska et al., 2009b).
Electron transfer is another form of non-radiative energy transfer that has been utilized to study the conformational movements of proteins. In electron transfer, the donor fluorophore transfers its excited state energy to an acceptor through a quantum mechanical transfer or exchange of an electron (Dexter, 1953; Moser and Dutton, 1992; Moser et al., 1992). This process allows the excited state fluorophore to return to the ground state without emitting a photon. Thus, the fluorophore's fluorescence is quenched. Like FRET, the probability that the donor will be quenched by the acceptor in electron transfer is steeply dependent on the distance between the two probes. However, the relationship shows an exponential dependence on distance not a sixth-power of distance-dependence like FRET (Lakowicz, 2006; Moser et al., 1992; Turro, 1978). Most electron transfer processes occur efficiently only over extremely short distances (<10 Å). In this regard, electron transfer is generally used to study very close interactions (5-15 Å) between two probes. For the study of proteins, tryptophan residues can participate in electron transfer reactions as either a donor or an acceptor. For example, electron transfer between tryptophan and bimane has been used to study the structure and movements of proteins (Islas and Zagotta, 2006; Mansoor et al., 2002; Semenova et al., 2009; Yao et al., 2006). In these studies individual cysteines were introduced into specific locations of receptors or channels and tryptophans were introduced into selected regions near these cysteines. The quenching of bimane-modified cysteines by these introduced tryptophans was then monitored with lifetime spectroscopy or patch-clamp fluorometry. When a tryptophan was nearby the lifetime or intensity of the bimane probe was significantly reduced. Furthermore, changes in distance between the tryptophan and the bimane were reflected in changes in bimane fluorescence. While electron transfer is useful for measuring proximity and changes in distance, it is limited in its ability to report absolute distances because of significant orientation dependence. Electron transfer between tryptophan and metal ions has also been used to study the structure of the synaptic protein alpha-synuculin (Lee et al., 2005; Lee et al., 2004). Also, electron transfer-induced quenching of tryptophan by histidines has been used to test for the opening and closing of the LacY sugar transporter (Smirnova et al., 2009). Finally, the non-fluorescent cysteine-reactive spin-label probe TEMPO has been shown to participate in electron transfer with fluorescent dyes (Zhu et al., 2005). This quencher exhibits an exponential dependence of quenching on distance and can be used as a short-range structural probe for protein structure.
For over 50 years fluorescence has been used to probe the structure of biological molecules. Its use is only expanding in prominence and utility. The advent of new microscopic techniques, fluorescent proteins, probes, and labeling strategies has rapidly expanded the fluorescence tool box for the study of protein structure and function. In particular, the development of both single molecule and live cell imaging has allowed the behavior of proteins to be studied in their native environment. These technical breakthroughs will continue to expand and enrich our understanding of molecular machines.
We would like to thank Alex Merz, Anne Carlson, Michael Puljung, and Noah Shuart for helpful comments on the manuscript. We would also like to thank Wolfhard Almers, Ehud Isacoff, and David Farrens, who kindly proved figures for reproduction. W.N.Z. was supported by the Howard Hughes Medical Institute (HHMI) and National Institutes of Health (NIH). J.W.T was supported by HHMI, the Jane Coffin Childs Foundation, and a NIH Pathway to Independence Award.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.