Allosteric mechanism in AMPA receptors: A FRET-based investigation of conformational changes

Abstract

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors are the primary mediators of fast excitatory synaptic transmission in the mammalian CNS. Structures of the extracellular ligand-binding domain suggest that the extent of cleft closure in the ligand-binding domain controls the extent of activation of the receptor. Here we have developed a fluorescence resonance energy transfer-based probe that allows us to study the extent of cleft closure in the isolated ligand-binding domain in solution. These investigations show that the wild-type protein exhibits a graded cleft closure that correlates to the extent of activation, which is in qualitative agreement with the crystal structures. However, the changes in extent of cleft closure between the apo and agonist-bound states are smaller than that observed in the crystal structures. We have also used this method to study the L650T mutant and show that in solution the α-amino-5-methyl-3-hydroxy-4-isoxazole propionate-bound form of this mutant exists primarily in a conformation that is more closed than predicted based on the activity, indicating that the degree of cleft closure alone cannot be used as a measure of extent of activation of the receptor, and there are possibly other mechanisms in addition to cleft closure that mediate the subtleties in extent of activation by a given agonist.

Ligand-gated ion channels are allosteric proteins that convert chemical signals into electrical signals by forming transmembrane ion channels upon ligand binding to an extracellular domain. Ionotropic glutamate receptors, a member of the ligand-gated ion channels, serve as an excellent paradigm for studying allostery in this family of proteins (1–6). The crystal structures of the isolated ligand-binding domain of the GluR2 subunit of the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor protein show a graded cleft closure with a direct correlation between the extent of cleft closure induced by a ligand and its extent of activation (5, 6), which is consistent with a multiple state-induced fit model. However, there is one exception to this correlation, the AMPA-bound form of the L650T mutation of the receptor crystallizes in two forms: one structure where the cleft is closed 11° and a second structure in which the cleft is closed 22° relative to the open apo form of the protein (7). These structures lie on either side of the cleft closure versus activation correlation observed for the wild-type protein, thus raising the question as to whether some ligands or modifications at the AMPA receptors can change the mechanism of activation from the multistate model to that defined by equilibrium of two distinct states.

Here we have developed a fluorescence resonance energy transfer (FRET)-based assay that allows us to investigate the conformational changes in the ligand-binding domain in solution, thus allowing us to probe the conformational changes in this domain without the crystallographic constraints. In addition, we have used this probe to determine whether the AMPA-bound form of the L650T exists in two distinct populations, consistent with the hypothesis that the intermediate activation by AMPA in this mutation is dictated by an equilibrium of a low-activity and high-activity state.

Results and Discussion

To perform FRET measurements, two strategies were used for attaching the donor and acceptor fluorophores. In one case, a histidine tag was introduced at the N terminus of the isolated ligand-binding domain of the GluR2 subunit (GluR2-S1S2) of the AMPA receptor, and the residue S652 in domain 2 was mutated to a cysteine (Fig. 1). These sites were tagged with a Cy3 derivative of nitrilotriaacetic acid chelate of nickel (ref. 8) acceptor and diethylenetriaminepentaacetic acid chelate of terbium (DTPA-Tb) (ref. 9) donor. In the second case, residues T394 (residue 5 in the GluR2-S1S2 protein) at the N terminus and S652 in domain 2 were mutated to cysteines, and these sites were tagged by using a 1:1 ratio of the maleimide derivatives of fluorescein (acceptor) and either DTPA-Tb or triethylenetetraaminehexaacetic acid chelate of terbium (TTHA-Tb) (Invitrogen) (donor). Although this type of 1:1 donor:acceptor labeling leads to a heterogeneous population of channels with different numbers of donors and acceptors, by using terbium chelate as the donor, which has a millisecond luminescence, only donor-acceptor pairs generate sensitized emission signals in the microsecond to millisecond time scale (delayed emission of the acceptor after receiving energy from the donor), and channels that contain all donors or all acceptors do not contribute to this signal.

Fig. 1.

Structure of GluR2-S1S2 showing the residues T394(C), and S652(C) as well as the Histag that are tagged with donor and acceptor fluorophores.

The inherent cysteines (425 and 436) were not mutated because they do not react under the conditions of the FRET experiments. The inacessibility of the cysteines was established by labeling the S1S2 protein with the S652C mutation with a 1:1 ratio of terbium chelate (donor) and fluorescein. As expected, the accessible 652 residue reacted with the fluorophores providing the donor only (milliseconds) and acceptor only lifetimes (nanoseconds), but no FRET lifetime was observed in the microsecond timescale (Fig. 2), thus establishing that no fraction of the protein had both donor and acceptor tags and, therefore, no accessible cysteine other than the S652C residue.

Fig. 2.

Donor only lifetime (A) and the lifetime measured at the acceptor wavelength (B) for S652C labeled with fluorescein and terbium-DTPA in the apo state showing that there is no long-lived lifetimes corresponding to FRET.

The fluorescence lifetimes for the donor terbium in the absence of an acceptor and the FRET lifetime as measured by the sensitized emission of the acceptor for all of the three constructs are shown in Fig. 3. For all of the three constructs, the FRET lifetime decreases upon going from the apo to the full agonist glutamate or AMPA-bound states, with the kainate-bound state having an intermediate lifetime (Fig. 3; see also Table 1, which is published as supporting information on the PNAS web site). On the other hand, no significant changes are observed in the lifetimes for donor only between apo, kainate-, glutamate-, and AMPA-bound states for all three constructs (Fig. 3 and Table 1). Based on these lifetimes, it is clearly evident that the FRET efficiency follows the order apo < kainate bound < glutamate bound ≈ AMPA-bound forms, therefore suggesting a decrease in the distance between the donor and acceptor in that order. These results are consistent with a cleft closure conformational change, with kainate inducing a smaller cleft closure relative to glutamate or AMPA, which is in qualitative agreement with the crystal structures.

Fig. 3.

Fluorescence lifetimes. The donor only lifetime (Left) and the FRET lifetime (Right) as measured by the sensitized emission for the apo (red), kainate-(blue), glutamate-(green), and AMPA-(black) bound forms of T394C-S652C labeled with fluorescein and DTPA-Tb (A), His tag-S652C labeled with (Nickel-NTA)₂-Cy3 and DTPA-Tb (B), and T394C-S652C labeled with fluorescein and TTHA-Tb (C).

The fluorophores are tagged at sites that reflect the cleft closure conformational change and are used to quantify the dependence of activation of the receptor on cleft closure. The absolute distances between the donor and acceptor fluorophores calculated from the FRET efficiency (Table 1) are plotted as a function of activation in Fig. 4A. The extent of activation was based on the maximum nondesensitizing currents mediated by the agonists normalized to the currents mediated by AMPA (Fig. 5). The distances for the four states for the two constructs, T394C-S652C tagged with fluorescein and DTPA-Tb and T394C-S652C tagged with fluorescein and TTHA-Tb, are virtually identical. This observation is consistent with the fact that the two constructs are similar and differ only in the structure of the chelate of the terbium donor. Additionally, the distances for the four states with the longer S1S2-histag construct are, as expected, larger than those for the T394C-S652C constructs. However, the changes in distances between the various ligated states for the three constructs show good agreement, with all of the constructs showing a 3-Å change between the apo and kainate state and a 5-Å change between the apo and AMPA states, consistent with the relative proximity of the tagged sites in the three constructs, which are expected to reflect the same large-scale cleft closure conformational change.

Fig. 4.

Dependence of cleft closure versus extent of activation. (A) Distance between T394C and S652C as observed in the crystal structure (filled squares) and FRET-based distances for T394C-S652C labeled with fluorescein and terbium-DTPA (open circles), T394C-S652C labeled with fluorescein and terbium-TTHA (filled triangles), and His tag-S652C labeled with (Nickel-NTA)₂-Cy3 and terbium-DTPA (open triangles), plotted as a function of extent of activation of the receptor as determined by the maximum nondesensitizing currents normalized to the currents mediated by AMPA. (B) Distance data from FRET measurements overlaid with those measured by crystal structures by using the AMPA-bound state as the reference.

Fig. 5.

Electrophysiological measurements of the L650V and wild-type protein. (A) The whole-cell currents mediated by 10 mM glutamate, 5 mM AMPA, and 5 mM kainate in the presence of 100 μM cyclothiazide with HEK293 cells expressing wild-type and L650V mutants. (B) The maximum currents obtained from at least three cells normalized to that mediated by AMPA for the wild-type, L650V, and normalized to that mediated by glutamate for the L650T mutant.

Although the trends in the change in distance between the N terminus and S652 as a function of activation are similar in both the FRET and crystal structures (Fig. 4A), with larger cleft closure leading to a larger activation of the receptor, crystal structures show changes of 3 Å and 7 Å for the changes between apo and kainate and apo and AMPA states, respectively. These differences in the distance changes are more evident when the FRET and crystal structure data are overlapped by using the AMPA-bound state as the reference (Fig. 4B).

The differences in distances between the crystal structures and FRET-based measurements, specifically between the apo and AMPA-bound forms, are significant and indicate that the crystal structures depict the extreme cases of open or closed forms and, in solution, a smaller change in cleft closure controls extent of activation. The smaller changes observed upon going from the apo to the ligated state in the FRET lifetimes are consistent with the small-angle scattering results. The scattering data for the apo state can be fit by a structure similar to the kainate-bound state (more cleft closure) or by an equilibrium of the apo and fully ligated crystal states (10). A kainate-like structure for the apo state would be consistent with the smaller changes in the distance between the apo and AMPA-bound forms observed in the FRET experiments described here.

To extend the structure-function correlations and to address the mechanism of activation of the L650T mutation by AMPA, the FRET-based probe also was used to study the conformational changes in the L650V and the L650T mutations (Fig. 6). The activations for the L650V, L650T, and wild-type protein are shown in Fig. 5, and the distance between S652 and T394 based on the FRET lifetimes (Table 1) are plotted as a function of activation in Fig. 7. The L650V mutant follows the same trends in the activation versus cleft closure as seen in the wild-type receptor with partial agonist kainate inducing a smaller cleft closure, suggesting a similar mechanism of allostery in this mutant as in the wild-type receptor. The L650T, on the other hand, does not follow the trends observed for the L650V and wild-type protein (Fig. 7). AMPA, a partial agonist of this mutant protein, exhibits larger cleft closure, whereas glutamate, which induces larger currents relative to AMPA, has a relatively more open cleft. Because the relative errors in the measurements are <2%, these differences are outside the errors associated with the FRET measurements.

Fig. 6.

The donor-only lifetime and the FRET lifetime as measured by the sensitized emission for the apo (red), kainate- (blue), glutamate- (green), and AMPA- (black) bound forms of L650V-T394C-S652C labeled with fluorescein and terbium-DTPA (A) and L650T-T394C-S652C labeled with fluorescein and terbium-TTHA (B). The residual for the single-exponential fit of the AMPA-bound L650T data also are shown at the bottom (brown).

Fig. 7.

The distances between T394C and S652C as measured by FRET for the wild-type (open squares), L650T (filled squares), and L650V (open triangles), and plotted as a function of extent of activation.

The FRET lifetime for the AMPA-bound form of the L650T is well represented with a single exponential that has a S652 and T394 distance closer to the crystal structure exhibiting a 22° cleft closure (Fig. 7). Because the distance, determined from the single-exponential fit of the FRET lifetime, represents the mean of a Gaussian distribution of distances between the donor and acceptor, it can be concluded that the AMPA-bound form of the L650T on average has a closed cleft conformation. Furthermore, because the FRET lifetime does not require a two-exponential fit, it can be concluded that two distinct populations with separated Gaussian distributions are not present in the AMPA-bound state of L650T. Thus indicating that the second conformation observed in the crystal structure of the AMPA-bound form of the L650T, exhibiting the 11° cleft closure, is not a major second conformation, and that the lower activation of AMPA in the L650T mutation is not due to an equilibrium of two distinct low- and high-activity conformations.

The FRET-based distances for the L650T mutation bound to agonist suggest that although cleft closure is required for channel activation in the mutant, the subtle differences in the extent of activation does not necessarily have to be controlled by the degree of cleft closure. Additionally, this result also implies that there have to be other mechanisms that couple the agonist-binding domain to the channel segments. One possibility is a direct coupling mechanism between agonist-binding domain and the channel segments through tertiary changes in the β-sheets in domain 2 of GluR2-S1S2 that are connected directly to the channel segments. This argument is strengthened by the observation that in the NMR structure of the glutamate-bound form of GluR2-S1S2, the dynamics of the β-sheet in the domain 2 of the receptor is in the timescale of receptor activation, suggesting that agonist binding could control channel activation through these secondary structures (11, 12). This type of a mechanism does not require the large-scale cleft closure to track the extent of activation and can account for the larger cleft closure observed in the AMPA-bound form of the L650T mutant.

Although the FRET-based probe shows that in some cases, such as the L650T mutant, the subtle differences in the extent of agonism may not exhibit a linear relationship between cleft closure and activation, for both the wild-type and mutant proteins, there is a clear distinction between the closed and activated states. Therefore, the FRET-based probes used here can serve as a high-throughput screen to identify ligands that are agonists or antagonists of this receptor. Such a probe would be particularly useful because agonists of the receptor have the potential to be used in Alzheimer’s and Parkinson’s diseases, and antagonists of this receptor are thought to be useful pharmacological agents in preventing neuronal degeneration associated with glutamate excitotoxicity in stroke and amyotrophic lateral sclerosis.

Materials and Methods

Fluorophores.

A Cy3 derivative of nitrilotriaacetic acid chelate of nickel was synthesized as outlined in ref. 8. The TTHA-Tb and DTPA-Tb (discontinued) were purchased from Invitrogen.

Protein Preparation and Fluorophore Labeling.

The plasmid pET22b(+)-GluR2S1S2J, used to express the GluR2-S1S2 protein, was kindly provided by Eric Gouaux (Oregon Health and Sciences University, Portland, OR). This construct has eight histidines at the N terminus, followed by sites for thrombin and trypsin digestion. For the His tag-S652C construct, the S652C mutation was introduced, and the histidine tag was not deleted during purification. All mutations were introduced by using the QuikChange site-directed mutagenesis kit (Stratagene). The mutations were confirmed by DNA sequencing. Proteins were expressed, purified, and characterized as described in ref. 7. In brief, the protein was expressed in Escherichia coli Origami-B(DE3) cells. After clarification and concentration, the cell culture supernatant was purified by a Ni-NTA HiTrap affinity column. For the T394C-S652C, L650V-T394C-S652C, and L650T-T394C-S652C, the histidine tag was removed by thrombin digestion after the purification step with the Ni-NTA column and further purified by using a SP-Sepharose column.

A 0.1–0.5 μM protein in phosphate buffer containing 1 mM glutamate was labeled with maleimide derivatives of terbium chelate and fluorescein (1:1) or with terbium chelate alone for the T394C-S652C, L650V-T394C-S652C, and L650T-T394C-S652C donor:acceptor and donor-only proteins, respectively. For the His tag-S652C protein, the protein was labeled with maleimide derivatives of terbium chelate for donor-only samples, and a 5 μM concentration of a Cy3 derivative of nitrilotriaacetic acid chelate of nickel was added for donor-acceptor samples. All of the constructs were established to be functional by characterizing the binding properties with tryptophan fluorescence (Supporting Methods, which is published as supporting information on the PNAS web site). The L650V and L650T mutations were made on the T394C-S652C background because the His-tag constructs of these mutants were found to be unstable and lost activity after a few days, even in the presence of ligands.

Electrophysiological Measurements.

For whole-cell current recordings, transfected HEK293 cells were voltage-clamped at a holding potential of −60 mV, and solutions were applied by using a homemade U-tube mixing device that had a 100-μm aperture. The electrode solution, for the electrophysiological measurements, contained 140 mM CsCl, 2 mM MgCl₂, 1 mM CaCl₂, 10 mM EGTA, 2 mM Na₂ATP, and 10 mM Hepes (pH 7.4); the extracellular bath solution contained 145 mM NaCl, 1.8 mM MgCl₂, 1 mM CaCl₂, 3 mM KCl, 10 mM glucose, and 10 mM Hepes (pH 7.4). Currents were amplified with an Axon 200B amplifier (Molecular Devices) and low-pass filtered at 1 kHz. The filtered signal was digitized by using a Labmaster DMA digitizing board controlled by pclamp (Axon Instruments, Foster City, CA). All of the experiments were performed at room temperature.

Fluorescence Measurements.

The fluorescence measurements were obtained by using a TimeMaster Model TM-3M/2003 (Photon Technology International, Lawrenceville, NJ), a cuvette-based fluorescence lifetime spectrometer. The source was a nitrogen laser that was fiber-optically coupled to the sample compartment. Emitted light was collected and passed through a monochromator to a stroboscopic detector. Data were collected by using felix 32 (Photon Technologies International, Lawrenceville, NJ) and analyzed by using origin (OriginLab, Northampton, MA). The lifetimes shown in Figs. 2 and 4 represent an average of three sets of data obtained on three different days. Each individual data exhibit the same trends in the lifetimes. The donor-only lifetimes were collected at 488 nm, whereas the FRET lifetimes were obtained by studying the sensitized emission of the acceptor, which was collected at the acceptor wavelength of 515 nm. The advantages of studying the FRET lifetimes at the acceptor wavelength instead of the donor wavelength (Fig. 8, which is published as supporting information on the PNAS web site) are outlined in Supporting Methods. The distances between the donor and acceptor were calculated by using Förster’s theory for energy transfer, and the R₀ values were determined for each of the constructs by obtaining the fluorescence and absorption spectrum of donor and acceptor respectively tagged to the proteins (Supporting Methods).

Abbreviations

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

DTPA-Tb

diethylenetriaminepentaacetic acid chelate of terbium

TTHA-Tb

triethylenetetraaminehexaacetic acid chelate of terbium.