Structural Dynamics of Glutamate Signaling Systems by smFRET

Abstract

Single-molecule Förster resonance energy transfer (smFRET) is a powerful technique for investigating the structural dynamics of biological macromolecules. smFRET reveals the conformational landscape and dynamic changes of proteins by building on the static structures found using cryo-electron microscopy, x-ray crystallography, and other methods. Combining smFRET with static structures allows for a direct correlation between dynamic conformation and function. Here, we discuss the different experimental setups, fluorescence detection schemes, and data analysis strategies that enable the study of structural dynamics of glutamate signaling across various timescales. We illustrate the versatility of smFRET by highlighting studies of a wide range of questions, including the mechanism of activation and transport, the role of intrinsically disordered segments, and allostery and cooperativity between subunits in biological systems responsible for glutamate signaling.

Introduction

Recent developments in structural biology, mainly through cryogenic electron microscopy (cryo-EM) techniques, have enabled the study of membrane proteins and their complexes faster than previously possible through x-ray crystallography (1, 2, 3). These “snapshots” of the proteins in multiple conformations provide a rich background for investigations of the conformational dynamics necessary to understand the mechanisms mediated by these proteins through a multitude of biophysical methods. Methodologies such as nuclear magnetic resonance (NMR) and electron paramagnetic resonance (EPR), which provide insights into the conformational dynamics, are challenged by the requirement of high concentrations of samples and limitations in protein size. Additionally, NMR and EPR provide a weighted average of the heterogeneous population, causing critical information about the individual dynamics and intermediate configurations to be lost. Single-molecule methods offer a unique solution to ensemble conditions by providing simultaneous structural and kinetics information from proteins in motion. The ability to observe protein conformation as a function of time allows for a direct study of the conformational transitions and kinetics between states (4).

Single-molecule Förster resonance energy transfer (smFRET) has become a mainstream technique for probing biomolecular structural dynamics. As the number of laboratories using smFRET increases, it becomes imperative to create a standardized technique to ensure proper usage. Furthermore, it is essential to ensure that the proper smFRET experimental design is used to answer the question of interest. Förster resonance energy transfer (FRET) uses the nonradiative transfer of energy from a donor fluorophore to a nearby acceptor fluorophore to accurately measure the distance between them (4,5) because the efficiency of the energy transfer depends on the inverse distance between fluorophores to the sixth power (6). However, selecting the appropriate fluorophores and their placement into biomolecules is of crucial importance. Various types of fluorophores have been employed for smFRET, including genetically encoded fluorophores like GFP (7,8) and organic fluorophore molecules like the cyanine dyes (9, 10, 11). Additionally, significant work has been done to develop new fluorophores for use in smFRET and to improve the performance of existing fluorophores (11, 12, 13). The most critical criteria to consider when selecting a fluorophore are the R₀ factor of the fluorophore pair, how the fluorophore will be attached to the site of interest within the protein, and the size of the fluorophore. The R₀ factor represents the distance in Angstroms between two fluorophores, at which those fluorophores experience half maximal energy transfer. Because FRET measurements are most sensitive to changes in distance when they are close to the R₀, the fluorophores being used should have an R₀-value that is close to the distance between the two sites being measured. Another important factor is the method by which the fluorophore will be attached to the protein of interest. Strategies for attaching fluorophores in a site-specific manner are discussed below. Additionally, the size of the fluorophore is significant because large fluorophores can experience steric hindrance with the protein of interest. Steric clashes can alter the conformation of the protein being studied and can also affect the relative orientation of the fluorophores, which can cause error in the measurements. Energy transfer depends upon the alignment of the fluorophores’ dipoles in relation to one another. This parameter, referred to as the dipolar orientation or κ², could be estimated by measuring the fluorescence anisotropy to minimize errors in the determined distances (14). For best practices, determining the κ² distribution for each FRET-labeled sample is recommended instead of the assumption of isotropic averaging, leading to the common use of 2/3. Moreover, the use of fluorophores with long linkers attached to proteins favors isotropic averaging compared with short-linker fluorophores, which restrict mobility. Therefore, longer linkers lead to small experimental errors to a maximum of 7% when considering all uncertainties in the measurements (15). Energy transfer is only possible when the emission spectrum of the donor overlaps with the excitation spectrum of the acceptor (5). Once compatible fluorophores are selected, interfluorophore distances from 10 to 136 Å, with up to 3-Å accuracy (15, 16, 17, 18), are measurable using mainly two modalities: ratiometric intensity-based methods or time-resolved fluorescence (19). By attaching compatible fluorophores to a protein, the distance between two locations on the protein is measurable at an instant in time. Special care must be taken to ensure that the incorporation of the fluorophores does not alter the protein (20,21). A variety of attachment chemistry exists to link fluorophores to proteins, including click reactions (22), cysteine reactions with electrophiles (23), and biorthogonal azide-alkyne chemistry via introduction of unnatural amino acids (24,25). Most of the time, the selection of the reactive chemistry is imposed by the wild-type sequence of the protein under study. For the purpose of studying membrane proteins, different treatments are compatible with smFRET experiments. Isolated membrane proteins can be solubilized in detergents (26). Alternatively, incorporating proteins into lipid-containing nanodisks (27) or reconstituted vesicles (28,29) can preserve the physiological lipid context of the protein. Live cell measurements are also possible (8,30).

smFRET implementations can be done with freely diffusing molecules or immobilized molecules on surfaces. Furthermore, data can be collected using confocal measurements or widefield imaging (Table 1). Selecting the appropriate combination depends on the question being asked. Typically, with diffusion-based experiments, confocal measurements are taken. In this case, labeled proteins of interest diffuse through the well-defined confocal volume of the observing microscope, and short bursts of photons are observed (31). These types of experiments are ideal when the dynamic behavior of interest is faster than the average traversal time through the confocal volume (Fig. 1). Furthermore, the concentration of fluorescently tagged protein within the sample must be low enough to ensure that only one fluorescent protein is within the confocal volume at a given time. If a higher protein concentration is required, one could fluorescently tag a portion of the proteins. This will help ensure only one FRET pair is present for a given measurement. In the case of immobilized molecules, molecules are linked to a coverslip with the most appropriate chemistries. Readers are encouraged to read (32, 33, 34) for more details. When immobilizing molecules, care must be taken to ensure that the fluorophores do not bind nonspecifically to the protein of interest or to the slide surface (10). Then individual molecules are located and measured over time using photon-counting detectors via confocal detection or using cameras (CMOS, EMCCDs) via widefield. These experiments are useful for observing individual molecules over longer temporal timescales. Photobleaching events are usually tracked to assure only a single molecule is observed at any given time.

Table 1.

Comparison of Confocal and Widefield for Freely Diffusing Molecules and Immobilized Molecule Measurements

	Confocal		Widefield
Molecule Freedom	Diffusing	Immobilized
Dynamic ranges	picoseconds to diffusion time	nanoseconds to minutes	submillisecond to minutes
Detectors	confocal, photon-counting modules (i.e., PMTs and APDs with TCSPC)		TIRF (widefield) EMCCDs and CMOSs
Number of molecules	limited diffusion through the sample volume, can be over 100,000	100–500a	limited by molecules in widefield, is ∼500
Data analysis	trace, burst-wise, BVA, fluctuation spectroscopy, and time resolved		trace, BVA, fluctuation spectroscopy, and camera or image based

APD, avalanche photodiode; PMT, photon-multiplier tube; TIRF, total internal reflection fluorescence.

^aTCSPC is less practical in immobilized molecules because of the lower number of sample molecules, which leads to lower photon-counting statistics sample size and limits the minimal dynamic range.

Figure 1.

smFRET investigations across different timescales. Top: the typical timescales for studies of various areas of glutamate signaling are shown as horizontal bars. The areas of study include both functional areas (e.g., amino acid transport) and structural areas (e.g., the extracellular domain (ECD)). ECD, TMD, and IDR refer to the various domains of the iGluRs as shown in the inset. smFRET investigations of glutamate signaling have focused primarily on the second-to-millisecond timescale, with submillisecond studies lagging behind. Bottom: various detection and analysis schemes exist for performing smFRET experiments at a variety of time resolutions. The ranges of timescales, across which the various methods are useful, are shown as horizontal arrows. Solid arrows indicate optimal timescales while dashed arrows indicate sub-optimal time scales. The timescale of diffusion of free-moving molecules is also shown as the timescale of the dynamics of interest can determine whether a diffusion-based or immobilized scheme is more useful for a planned investigation of single-molecule dynamics. To see this figure in color, go online.

It is imperative that experiments be designed with the appropriate temporal scale in mind (Fig. 1). There are three significant factors in smFRET experiments that affect the achievable temporal resolution: protein treatment, detection mode, and data analysis (4). It has already been mentioned that proteins can be immobilized or allowed to diffuse freely. After determining which of these methods works best for making observations at the right timescale for the dynamics of interest, the proper detector must be selected. It is typical to immobilize proteins and use a widefield camera when measuring conformational dynamics that are slower than milliseconds. The photon-to-electron conversion of most widefield camera detectors, such as common charge-coupled devices (CCDs), electron-multiplying CCDs (EMCCDs), and complementary metal-oxide semiconductors (CMOSs), limits the temporal resolution to be around tens of milliseconds (Fig. 1). This temporal resolution is suitable for measuring larger domain motion and large-scale conformational rearrangement of proteins (32). By binning pixels, CMOS cameras can increase their temporal resolution with a time resolution of up to 250 μs being reported (35), but this lowers the spatial resolution. Photon-counting modules like avalanche photodiodes (APDs) and photon multiplier tubes (PMTs) can be utilized to further increase the temporal resolution. Temporal resolution is going to be based on the data-acquisition system used, i.e., Data Acquisition boards (DAQ) and Field-Programmable Gate Arrays (FPGAs). Because these modules cannot differentiate between emission wavelengths, it is crucial to ensure the correct filters are being used based on the selected fluorophores. Photon counting allows the measurement of faster dynamics such as local flexibility and side-chain rotations with a temporal resolution ranging from 10 μs to 100 ms (36), including several orthogonal analysis methods, that increases the extracted dynamics and the temporal information obtained from smFRET measurements (37, 38, 39). Time-correlated, single-photon-counting (TCSPC) electronics will measure fluorescence and luminescence lifetimes ranging from nanoseconds to microseconds, respectively (38). Fluorescence correlation spectroscopy (FCS) and its different variations allow conformational dynamics to be determined over a broader timescale (40, 41, 42, 43).

There exist many different methods for smFRET data analysis, depending upon the information gathered. Trace analysis is typically modeled with Hidden Markov modeling (HMM) (44,45) and many variations (46). Burst-wise analysis, which looks at the pattern in photon detection to differentiate donor from acceptor in confocal measurements (47), is often done by probability distribution analysis (PDA) (48) or burst variance analysis (BVA) (49). The choice of data analysis depends upon the data collected and experimental design. If FCS and lifetime information is gathered, these data should be included in data analysis because it will only help enhance the results. Multiparameter fluorescence detection (MFD) (19,50) will ensure the best results are obtained. Many laboratories are actively developing new software and setting up best practices in terms of data analysis and offer these packages online for free. These community-based resources range from online software-sharing hubs such as the FRET community (51) and the kinSoftChallenge (46) to multilaboratory studies of FRET precision (15). Given that the size of the protein of interest and the sample concentration are not limitations for FRET experimental design, smFRET is a versatile methodology to study membrane proteins and has been used in a wide range of systems such as G-protein-coupled receptors (GPCRs), transporters, and ion channels. Here, we choose the three systems related to glutamate signaling as an example system in which smFRET methods have provided invaluable biophysical insights. This system was chosen because it has the common theme of glutamate signaling but with three unique systems that covers GPCRs, transporters, and ion channels.

Glutamate signaling systems studied by smFRET

Glutamatergic signaling is the primary form of excitatory signaling in the mammalian central nervous system. In this type of signaling, the neurotransmitter glutamate is released into the synapse, where it binds to and activates various receptor molecules, including ion channels and metabotropic receptors. After this glutamate release, the glutamate must undergo reuptake into the presynaptic neuron and be packaged into neurotransmitter vesicles, a process facilitated by amino acid transporter molecules (46). smFRET studies have been carried out on proteins from each of these three families of proteins that are involved in glutamatergic signaling. Although the metabotropic glutamate receptor (mGluR) and ionotropic glutamate receptor (iGluR) share some similarities in the glutamate binding site, the transmembrane segments and signaling processes are very different, one being G-protein coupled and the other an ion channel. The glutamate transport has a completely different architecture both at the ligand binding site and in the transmembrane transporter segments. Thus, although the three systems chosen are linked by the fact that they are involved in glutamate-mediated signaling, the wide range of architectures illustrates the versatility of smFRET as a tool to study the dynamics and conformations of a wide variety of proteins.

One family of proteins that are involved in the glutamatergic signaling process are the mGluRs, which are class C GPCRs with large bi-lobed extracellular domains that bind to the agonist glutamate. Based on end-state structures of agonist- and antagonist-bound forms of the extracellular domain, the mechanism for activation by agonists was suggested to be a closure of the bi-lobed cleft (52, 53, 54, 55, 56). smFRET investigations on the soluble extracellular domain that utilized a diffusing experimental setup with confocal detection were able to put dynamics into the context of activation and showed that the domain rapidly fluctuates between the open and closed cleft states and that the extent of stabilization of the protein in the closed cleft state dictated the extent of agonism (Fig. 2 A; (57)). Although later smFRET measurements on the full-length receptor that analyzed immobilized molecules using a TIRF-based detection scheme showed three major states when measuring across the dimeric extracellular domain at the same sites as in the soluble domain, agonist efficacy was still shown to be dictated by the occupancy of a closed cleft state that is consistent with the smFRET studies on the isolated extracellular domain (58). However, the dwell times for interconversion between the FRET states were found to be in the range of tens of milliseconds and not as rapid as those observed for the isolated extracellular domains. It is possible that the slower dynamics is due to the stabilization and restricted dynamics of the extracellular domains in the presence of the transmembrane domains, showing the importance of investigating the dynamics in the full-length systems.

Figure 2.

smFRET measurements of proteins involved in glutamate signaling. The locations of smFRET fluorophore labeling sites and measurements are shown. These sites are chosen with the goal of studying established (based on previous structures) or hypothesized conformational changes in these proteins. (A) Shown are the measurements performed on the agonist-binding dimer of mGluRs (57, 58, 59, 60) (Protein Data Bank, PDB: 1EWT (54)). This measurement site reports on the conformational changes at the interface between the mGluR dimer caused by the closure of the bi-lobed cleft upon binding agonists. (B) Shown are the measurements performed on the trimeric glutamate transporter (61, 62, 63) (PDB: 1XFH (64)). These measurements report on the distance between adjacent monomers of the transporter trimer that are expected to monitor motions associated with transport. (C) Shown are the measurements performed on an iGluR (26) (PDB: 6MMK (65)). The sites across the agonist binding domain monitor motions across the bi-lobed cleft caused by agonist or antagonist binding. The sites across the transmembrane segments are expected to monitor motions across the ion pore. The side chains of fluorophore-labeled residues are shown as spheres, whereas distances being investigated are shown as dotted lines. To see this figure in color, go online.

smFRET measurements of immobilized molecules were also used to investigate the role that conformational changes play in determining the cooperativity between the agonist binding sites in mGluRs (59). These studies showed that agonist-induced closure of one cleft leads to an allosteric shift in the dynamic equilibrium of the second unliganded subunit in the dimeric mGluRs. Thus, the higher spontaneous basal dynamics and closure of the mGluR3 influenced the liganded mGluR2, leading to higher activation in the heteromer. This mechanism of cooperativity was found to be universal and translated to mGluR2/mGluR7 heteromers, highlighting the importance of such heteromerization in physiology (60). mGluR7 homodimers have a significantly lower affinity and activation relative to the other mGluRs. However, in the context of mGluR2/mGluR7 heteromers, both efficacy and affinity are high. smFRET studies showed that the heteromeric receptor exhibits faster state-to-state transitions and, therefore, has increased conformational dynamics relative to the homomeric receptor. Additionally, synthetic agonists selective for either mGluR2 or mGluR7 produced a more substantial shift to low FRET (related to higher cleft closure at the agonist binding domain) than observed when both subunits were bound in the corresponding homomeric receptor, correlating well to a stronger activation. These experiments that allow for studying individual heteromeric combinations are possible because specific combinations can be pulled down onto the slides used for smFRET imaging (10,26,66). This selective pulldown of heteromeric receptors is achieved by placing the fluorophore-labeling sites on one subunit and placing the affinity tag for attachment to the slide on the other subunit. Using this strategy, molecules that do not contain a subunit with the affinity tag will not attach to the slide, and molecules that do not contain any fluorophore-labeling sites will not be observed. This ensures that immobilized molecules that exhibit an FRET signal contain at least one of each of these two subunits.

Glutamate transporters catalyze neurotransmitter uptake from the synaptic cleft into the cytoplasm of glial cells and neurons. The bacterial homolog of the glutamate transporter, the sodium-aspartate symporter from Pyrococcus horikoshii, Glt_pH, has served as a model system for studying structure-function correlations in this family. The initial framework provided by the x-ray structures of the protein in several conformations along with the slow turnover rate of tens of seconds for this protein made it ideal for direct structure-function studies using tethered smFRET (Fig. 2 B; (61,62,64,67)). These studies provided direct evidence for “elevator-like” motions for transport and also showed that each subunit could undergo motions independent of each other (28,62,63). This motion also has a burst-like pattern with periods of quiescence and periods of rapid transitions (28). The rapid dynamic mode was hypothesized to be due to the separation of the transport domain from the trimeric scaffold, allowing for the rapid domain movements across the bilayer. Two mutations introduced into Glt_pH for imparting characteristics of the human glutamate transporter lead to increased transport domain dynamics. The increased dynamics correlated to the increased rate of substrate transport observed in the human glutamate transporters relative to Glt_pH, providing a direct temporal relationship between transport domain motion and substrate uptake (63). Several of these investigations were performed in proteoliposomes, with the transporter tethered on the slide. Tethering allows for a specific outside-out orientation of the transporter, making it possible to tune the concentration of luminal and external ions and substrates, leading to more physiologically relevant studies.

An additional class of membrane proteins that are investigated through smFRET is ion channels. Extensive smFRET investigations have been carried out on members of the iGluRs. Initial studies were performed on the soluble agonist binding domain of the receptor, in particular the AMPA subtype of the iGluRs. X-ray structures of the isolated agonist binding domain showed a correlation between the extent of cleft closure at the bi-lobed agonist binding domain and the extent of activation in several cases (68, 69, 70, 71, 72, 73). However, this relationship broke down in mutants, such as the T686S, in which the cleft showed full closure even when the extent of activation was only partial (74). smFRET measurements that utilized immobilized protein molecules on a surface probed all the states that the protein occupies and reconciled this issue by showing that the mutant protein occupied a wide range of conformations and that the activation is dictated by the fractional occupancy of the high FRET closed cleft agonist binding state (75,76). More importantly, the smFRET studies also showed that the distance changes between the most probable states for the different liganded conditions and the mutant correlated with the x-ray structures, showing that distance changes can be accurately measured by smFRET.

Similarly, cryo-EM structures of full-length AMPA receptors showed varying degrees of decoupling across the dimers within the tetrameric amino-terminal domain of the receptor that is associated with receptor desensitization. Hence, the role of this decoupling in desensitization was debated (77, 78, 79). The complete conformational landscape that could be probed with smFRET at this site using the immobilized full-length receptor showed that largely decoupled states did exist under desensitizing conditions. However, a larger fraction of the receptor showed smaller decoupling, this large decoupling is not required for desensitization (80). Additionally, based on the distances, we could directly relate the conformations observed in the smFRET data to specific cryo-EM structures (80). Recently, smFRET investigations were used to study the mechanism of cooperativity between the agonists glutamate and glycine in the NMDA receptor subtype of the iGluRs (Fig. 2 C; (26)). The smFRET investigations of immobilized receptors showed that the binding of one agonist causes a stabilization of the closed cleft bi-lobed agonist binding domain and lower conformational fluctuations at the site where the agonist binds but an increase in conformational flexibility and dynamics at the second agonist site. The loss of such an effect in a mutant receptor where such negative cooperativity was not observed (81) confirmed that certain conformational states observed via smFRET at the second agonist binding site contribute to the lower affinity of the second agonist when the first agonist is bound to the receptor. These studies again highlight the importance of conformational dynamics in function along with the need to understand the complete conformational landscape of the receptor that is possible through single-molecule methods.

smFRET also stands to contribute significantly to our understanding of macromolecular dynamics for intrinsically disordered proteins (IDPs) or intrinsically disordered regions (IDRs). The inherently dynamic nature of IDPs is incompatible with analysis for static structure through cryo-EM or x-ray crystallography, either because the study is unsuited for analyzing a molecule with little structure or because the results fail to capture a complete understanding of the dynamics of the protein. However, smFRET can report on the conformation of a protein over time without averaging over an ensemble. Combining static measurements and smFRET allows the gathering of valuable information concerning IDPs or IDRs (82,83). For example, from studies of both diffusing and immobilized proteins, it was determined that the IDR C-terminal domain (CTD) of the GluN2B subunit is required for GluN2B to regulate NMDARs (57,60,84, 85, 86). The use of smFRET has helped to understand the role of posttranslational modifications on the iGluRs’ disordered intracellular segments (87, 88, 89). Bowen and Choi used smFRET measurements of immobilized molecules to show differences in conformational dynamics as well as changes in the extent of disorder due to phosphorylation (87,90). Given that there is little structural insight for this segment of the protein, these early studies pave the way for future investigations into how disorder and dynamics in these segments modulate receptor function.

Perspective

To observe several essential conformations for mediating functions of glutamate singling systems, smFRET must be used. This allows for determining the complete conformational landscape and dynamics that are not observed using only structural methods, such as cryo-EM and x-ray crystallography. smFRET also highlights the dynamic nature of these systems and the role of such dynamics in function. The smFRET technique’s versatility in terms of the ability to probe protein complexes of all sizes with minimal requirements in terms of concentration and in a near-native state allows for similar investigations on the large number of membrane proteins for which end-state structures are available through cryo-EM. Moreover, with the ability to measure FRET with high precision, it is now possible to use FRET-derived distances in combination with other structural biology tools or molecular dynamics (MD) simulations to model protein structures (91,92) and deposit them in pdb-dev (93). Static structures enable the rational design of starting points for MD simulations and the selection of fluorophore attachment points for smFRET measurements. MD simulations show dynamic motions not revealed with static structures alone and reveal the regions of a molecule that will show biologically relevant motions that can be measured with smFRET. smFRET measurements can provide constraints for MD simulations that enhance the physiological relevance of simulation results and provide insight into highly dynamic regions of proteins that cannot be well resolved through static structural methods alone. However, it is important to understand all the physical aspects and parameters of the dye to obtain accurate simulations. As machine learning continues to advance, it will become a great tool for smFRET data analysis (94,95).

Furthermore, the field of correlative microscopy is yet to expand on the possibility of combining FRET and cryoEM. In this way, structural biology techniques, MD simulations, and smFRET measurements can be combined to provide a greater understanding of various biological systems on the single-molecule level. The development of detectors and data analysis methods that function at submillisecond time resolutions has enabled the investigation of new research questions relating to the structure and dynamics of individual molecules. However, the studies that utilize smFRET have focused primarily on the timescale of seconds to milliseconds, with submillisecond studies lagging (Fig. 1). The future of the smFRET field will involve expanding investigations into the submillisecond dynamics of individual molecules. As the tools and methods needed to investigate submillisecond dynamics become more widely available, with data analysis methods becoming increasingly standardized, the area of submillisecond dynamics will continue to expand and answer as-yet-unresolved research questions.

Author Contributions

R.J.D. wrote the outline with guidance from H.S. and V.J. All authors contributed to writing the manuscript.

Editor: Meyer Jackson.