α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic Acid (AMPA) Receptor Channels Lacking the N-terminal Domain*

Ionotropic glutamate receptor (iGluR) subunits contain a ∼400-residue extracellular N-terminal domain (“X domain”), which is sequence-related to bacterial amino acid-binding proteins and to class C G-protein-coupled receptors. The X domain has been implicated in the assembly, transport to the cell surface, allosteric ligand binding, and desensitization in various members of the iGluR family, but its actual role in these events is poorly characterized. We have studied the properties of homomeric α-amino-3-hydroxy-5-methylisoxazolepropionate (AMPA)-selective GluR-D glutamate receptors carrying N-terminal deletions. Our analysis indicates that, surprisingly, transport to the cell surface, ligand binding properties, agonist-triggered channel activation, rapid desensitization, and allosteric potentiation by cyclothiazide can occur normally in the complete absence of the X domain (residues 22–402). The relatively intact ligand-gated channel function of a homomeric AMPA receptor in the absence of the X domain indirectly suggests more subtle roles for this domain in AMPA receptors, e.g. in the assembly of heteromeric receptors and in synaptic protein interactions.

Ionotropic glutamate receptors (iGluR) 1 mediate the majority of fast excitatory neurotransmission in the mammalian central nervous system. These ligand-gated channels are believed to be tetramers consisting of subclass-specific sets of homologous subunits, each ϳ900 -1300 amino acid residues in length (1,2). Previous studies have identified three segments in iGluR subunits that show sequence similarity to bacterial proteins involved in the transport of extracellular solutes (3). Two segments, S1 and S2, separated by the membrane-associated channel domain, form a two-lobed agonist-binding domain, which is homologous to bacterial polar amino acid-binding proteins (4 -7). In addition, an N-terminal ϳ400-residue segment preceding S1 shows sequence similarity to bacterial binding proteins specific for leucine (LBP) and leucine/isoleu-cine/valine (LIVBP) and to the N-terminal extracellular Nterminal segments of Class C G-protein-coupled receptors, which include the metabotropic glutamate receptors (3). The three-dimensional structure of this domain has not been determined, but homology models predict an LIVBP-like structure consisting of two lobes surrounding a central cleft (8,9).
The N-terminal domains of human and rat AMPA receptor subunits are ϳ95% identical. This high degree of sequence conservation is certainly indicative of an essential physiological function. In contrast to the agonist-binding domain and the transmembrane channel region, however, the functional role of this domain, which in AMPA and kainate receptors may represent up to 45% of the mature polypeptide, is poorly understood, hence it is referred to as the "X domain." In principle, functional activities assigned to or suggested for the N-terminal domain fall into three different classes. First, this domain is implicated as a determinant in the assembly of oligomeric channels (10 -12). Second, the X domain may mediate the allosteric transitions involved in the channel activation, desensitization, or modulation by ions and drugs as has been observed for NMDA receptors (8,(13)(14)(15)(16)(17). Third, the X domain may provide docking sites for extracellular proteins, which serve to cluster the receptors or stabilize their localization. In support of the latter possibility is a recent report of an extracellular interaction of AMPA receptors with a synaptic protein Narp, although the detailed site of the interaction was not localized (18).
To gain insight into the function of the N-terminal domain in the AMPA receptors, we have examined the properties of GluR-D (GluR4) receptors carrying N-terminal deletions. We report that, surprisingly, homomeric AMPA receptors that lack the entire X domain form functional glutamate-gated channels in transfected HEK 293 cells. A closer analysis of the properties of the mutant channels shows that the assembly of homomeric receptors, transport to cell surface, ligand binding, agonisttriggered channel activation, and rapid desensitization can occur normally in the complete absence of the X domain.

MATERIALS AND METHODS
DNA Constructs-Standard molecular biological techniques were used to clone the cDNA encoding residues 22-902 of rat GluR-D (flip isoform; residues 1-21 code for the signal peptide; Ref. 19) into a derivative of pcDNA3.1(Ϫ) (Invitrogen), which carries a viral signal peptide followed by an N-terminal FLAG epitope (6). All N-terminal deletion constructs were built on this vector (wild-type GluR-D). The deletion mutants, created by PCR methodology, were used to replace the native region in wild-type GluR-D by appropriate restriction enzyme digest. All PCR-generated constructs were verified by sequencing.
Expression in Insect Cells-The recombinant baculovirus for the expression of GluR-D ⌬402 was prepared by using Bac-to-Bac system (Invitrogen) according to manufacturer's instructions. High Five insect cells (Invitrogen) cultured in T75 flasks in serum-free SF-900II medium were infected with recombinant baculoviruses for FLAG-tagged fulllength GluR-D (11) and ⌬402 and harvested 48 h later.
Ligand Binding Analysis-Preparation of membranes, determination of [ 3 H]AMPA (specific activity 60.0 Ci/mmol; PerkinElmer Life Sciences) binding by using a filtration assay, and displacement experiments using unlabeled compounds were performed as described previously (11). Ligand binding data were analyzed by using the PRISM nonlinear curve-fitting software (GraphPad Inc.).
Expression in HEK 293 Cells and Immunofluorescence Staining-HEK 293 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and 2 mM L-glutamine and 1% penicillin-streptomycin solution. Immediately prior to transfection they were replated at a density of 2 ϫ 10 5 cells/ml. Transfection was done by using the calcium phosphate method, and the medium was changed the following day and the cells harvested 40 h following transfection. For patch clamp experiments the cells were cotransfected with a plasmid encoding green fluorescent protein (pEGFP-C1, Clontech) for visualization with epifluorescence microscope.
For immunofluorescence, cells were fixed in 3% paraformaldehyde, and preblocked with 3% goat serum. For total staining, the cells were additionally incubated in 0.05% Triton X-100 in phosphate-buffered saline prior to blocking step. Cells were labeled with monoclonal anti-FLAG IgG (M1, Sigma; 5 g/ml) followed by Cy3-conjugated anti-mouse IgG secondary antibody (Jackson Laboratories; 7 g/ml). Cells were examined using an Olympus Provis AX70 epifluorescence microscope and pictures collected by a Photometrics SenSys air-cooled CCD camera using Image ProPlus software.
Quantitative Immunolabeling-Transfected HEK 293 cells were plated in 96-well tissue culture plates coated with Matrigel (BD Biosciences) at a density of 4 ϫ 10 4 cells per well. Cells were fixed 40 h post-transfection by either 3% paraformaldehyde in phosphate-buffered saline (for surface labeling) or by 100% methanol at Ϫ20°C (for total labeling). Following incubation in 3% goat serum to block nonspecific binding, the cells were labeled with monoclonal anti-FLAG antibody (M1; 100 ng per well) for 1 h at 37°C. Cells were washed three times with 1% goat serum, and incubated with Eu 3ϩ -conjugated anti-mouse IgG (Wallac; 25 ng per well) for 1 h at 37°C. Thereafter, the cells were washed as above, rinsed with Tris-buffered saline, and Enhancement solution (Wallac; 100 l per well) was added to form the fluorescent Eu 3ϩ -chelate. The samples were measured on a Wallac VICTOR 2 instrument using excitation filter at 340 nm and emission filter at 615 nm settings and a delay time of 400 s. The fluorescence values (recorded as cpm) for the surface and total labeling were corrected for the nonspecific background obtained from HEK 293 cells transfected for expression of non-FLAG-tagged GluR-D.
Rapid applications of agonists were made using the SF-77B perfusion fast step system with a narrow-mouthed theta tube (Warner Instrument Corporation). Rise time of glutamate currents was less than 1 ms. No currents were observed in untransfected cells or cells transfected with a nonrelevant cDNA construct. The currents were recorded using an EPC 9 patch clamp with Pulse and Pulsefit software (HEKA Electronics). Desensitization rate constant ( des ) was determined by fitting a single exponential to the decaying component of currents elicited by 5 mM glutamate. Concentration-response relationships were fitted with a logistic equation using Microcal Origin software. The data were compared using Student's t test. All data are given as mean Ϯ S.E.

RESULTS
To gain insight into how the N-terminal X domain contributes to the properties of AMPA receptors, we analyzed receptors mutated to eliminate all or parts of this domain. N-terminally FLAG-tagged GluR-D (flip isoform) and mutants lacking increasing lengths of the N-terminal segment (residues 22-541; signal peptide: 1-21; Fig. 1A) of GluR-D were transiently expressed in transfected HEK 293 cells. The expression of the mutant proteins was verified by immunoblotting, which showed the presence of FLAG-immunoreactive bands with electrophoretic mobilities consistent with the sizes inferred for the deletion constructs from the sequence and predicted glycosylation status (Fig. 1B). The expression levels of the N-terminal deletion constructs ⌬162, ⌬303, and ⌬366 were roughly equivalent to that of the full-length GluR-D, but the two smallest constructs, ⌬402, lacking the entire X domain, and ⌬541, which lacked the whole N-terminal extracellular domain (i.e. both the X domain and the ligand-binding S1 segment), were expressed often at a somewhat lower level (Fig. 1B).
The subcellular distribution of the deletion mutants was determined by immunofluorescence microscopy. Under permeabilized conditions, all constructs showed intracellular staining by anti-FLAG antibody (Fig. 2). In contrast, only two constructs, the full-length GluR-D and ⌬402, devoid of the entire X domain, were expressed on cell surface as indicated by specific immunostaining under nonpermeabilizing conditions (Fig. 2). Both stainings had a finely punctate appearance that was evenly distributed over the cell membrane. These findings imply that deletions within (the predicted) structural domains (X and S1S2) of the receptor, unlike a complete deletion of the X domain, block the transport of the receptor to cell surface, possibly due to incorrect folding. The latter possibility is consistent with our finding that none of the deletion mutants was stained by a monoclonal antibody Fab22, which recognizes a conformation-sensitive epitope within the X domain of GluR-D (20), although cells expressing the full-length GluR-D were intensely stained both intracellularly and on cell surface (results not shown).
To complement the immunofluorescence studies, we used europium-labeled secondary antibody to quantitate the relative amounts of FLAG-tagged full-length GluR-D and ⌬402 present on cell surface. Time-resolved fluorescence measurements indicated that 84 Ϯ 14% (mean of surface labeling/total labeling Ϯ S.D.; n ϭ 4) of the full-length receptor and 66 Ϯ 13% of ⌬402 (n ϭ 4) were accessible to Eu 3ϩ -labeled anti-mouse IgG in fixed cells under nonpermeabilizing conditions; the slight decrease is statistically not significant (in nonpaired t test). No significant surface labeling was obtained with receptor constructs that were negative in immunofluorescence microscopy (results not shown). The expression level of ⌬402 was similar (105 Ϯ 8%) to that of the full-length GluR-D as determined by europium labeling.
Next, we wished to determine whether the presence or absence of the X domain has any influence on the interaction of the receptor with agonists and antagonist as determined by radioligand binding experiments. To obtain larger amounts of receptor protein to facilitate more accurate radioligand binding measurements, we used the baculovirus expression system to produce full-length GluR-D and GluR-D ⌬402. Virally infected High Five insect cells produced GluR-D and GluR-D ⌬402 as 110-kDa and 50-kDa FLAG-immunoreactive species (Fig. 3A). The unaltered membrane localization and ligand binding properties of ⌬402 suggested the possibility that functional receptors may be formed by GluR-D subunits lacking the X domain. To investigate this, outside-out patches were pulled from HEK 293 cells transiently expressing the full-length GluR-D and ⌬402. As expected, both L-glutamate and AMPA triggered rapidly desensitizing currents in GluR-D-containing patches. Surprisingly, however, highly similar agonist responses were recorded from the ⌬402-containing patches (Fig.  4A). The glutamate sensitivity of ⌬402 channels was also similar to that of wild-type channels (Fig. 4A). However, further analysis revealed a slight change in desensitization kinetics: the rate of desensitization of glutamate responses was somewhat slower for the ⌬402 channels ( des 4.52 Ϯ 0.31 ms, n ϭ 5) than for GluR-D (3.47 Ϯ 0.24 ms, n ϭ 6; p ϭ 0.024; Fig. 4B). Responses of ⌬402 channels to 1 mM glutamate were also fully blocked by the specific AMPA receptor antagonist CNQX (100 M; Fig. 4C), consistent with its unchanged antagonist binding characteristics.
Finally, we studied the effect of cyclothiazide, an allosteric modulator of AMPA receptor desensitization, on GluR-D and ⌬402. By using a filtration assay, we measured the inhibitory effect of cyclothiazide on [ 3 H]AMPA binding to GluR-D and to ⌬402 expressed on insect cell membranes. As shown in Fig. 5A, cyclothiazide caused a concentration-dependent inhibition of . These findings demonstrate that the cyclothiazide binding site, and the structures mediating its allosteric effect on agonist binding, are present in the deletion construct as well. We then studied the effect of cyclothiazide current responses of GluR-D and ⌬402 channels expressed in HEK 293 cells. Consistent with the ligand binding data, glutamate responses recorded from outside-out patches indicated that cyclothiazide removed desensitization of glutamate currents of GluR-D and ⌬402 channels in a similar manner (Fig. 5B). These findings clearly show that the ⌬402 construct contains a functional cyclothiazide binding site, thus demonstrating that the X domain is not a critical participant in the allosteric transitions involved in homomeric GluR-D receptor desensitization or its regulation by cyclothiazide. DISCUSSION Based on biochemical, cell biological, and electrophysiological analyses of native and recombinant iGluR, a number of different activities or potential physiological roles for the Nterminal X domain have been suggested. Our present analysis on the properties of GluR-D⌬22-402 receptors clearly exclude an essential role of the X domain in the assembly of homomeric channels, transport to cell surface, channel gating, rapid desensitization, and allosteric potentiation by cyclothiazide in AMPA receptors. It must be noted, however, that our study focuses on one AMPA receptor subunit, and the specific functions of the X domain are likely to vary among receptor subclasses, and possibly even among subunits. In the following, we discuss the possible functions of the N-terminal domain of GluR-D AMPA receptor in the light of the present results.
Channel Gating-Interestingly, the N-terminal domain is not present in the small kainate-binding proteins expressed in avian and amphibian species, which are otherwise homologous to iGluR subunits. These proteins do not seem to have a capacity to form active ligand-gated channels, even though their M1-M3 segment is able to mediate an ion conductance when artificially inserted into AMPA or kainate receptor subunits (22). These findings raise the possibility that the N-terminal domain might be essential for channel activity, possibly by participating in the conformational coupling between the ligand-binding domain and the channel domain. The identification of a prokaryotic glutamate-sensitive potassium channel ("GluR0"), which is distantly related to the core of iGluR sub-

FIG. 4. Functional characterization of GluR-D and ⌬402 receptors in outside-out patches pulled from HEK 293 cells. A, both
wild-type GluR-D and ⌬402 receptors are similarly activated by AMPA receptor agonists glutamate and AMPA. Holding potential was Ϫ60 mV in all experiments shown. The glutamate concentration response curves were also similar for GluR-D and ⌬402 receptors (EC 50 ϭ 2.5 mM, Hill coefficient ϭ 3.6 for wild type, and EC 50 ϭ 2.9 mM, Hill coefficient ϭ 3.9 for ⌬402). B, ⌬402 glutamate currents desensitize more slowly than GluR-D currents in response to 5 mM glutamate. The single exponential fitted to the recordings is drawn on top of original traces in black. The panel on the right shows mean desensitization time constants calculated from wild-type GluR-D (t des ϭ 3.5 Ϯ 0.2 ms, n ϭ 6) and ⌬402 (t des ϭ 4.5 Ϯ 0.3 ms, n ϭ 5) receptors. C, the AMPA receptor antagonist CNQX blocks the current elicited by 1 mM glutamate in cells expressing ⌬402. units and has no X domain, demonstrates that a direct coupling between the ligand-binding S1-S2 domain and the ion channel gate can take place without an obligatory involvement of an N-terminal domain (23). The overall sequence similarity between the prokaryotic channel and eukaryotic iGluR is fairly low, however, and even the ligand binding mechanism employed by GluR0 seems to profoundly differ that observed for the AMPA receptor (24). Therefore, it is entirely possible that the mechanism of ligand-triggered activation of the GluR0 channels is also fundamentally different. However, our present results convincingly demonstrate that in a eukaryotic AMPA receptor, the neurotransmitter-binding domain can directly couple to and regulate the ion channel without an involvement of the N-terminal domain.
Desensitization and Allosteric Modulation-Several studies have implicated the N-terminal domain of NMDA receptor subunits as a determinant or direct participant in some forms of desensitization. Glycine-independent desensitization of NMDA receptors was assigned to two regions in the NR2A subunit, the X domain and an area within the S1 segment (13,17). Moreover, participation of the N-terminal domain in the regulation of NMDA receptor channel function by allosteric modulators, including Zn 2ϩ ions (14,16), ifenprodil (8,15,16), spermine (8), protons (25), and redox agents (26), has been demonstrated. In some cases, this modulation has been shown to involve an direct allosteric interaction between the N-terminal domain and the agonist-binding domain (16).
So far, no studies have indicated any direct engagement of the X domain with allosteric modulation or desensitization of AMPA receptors. Indeed, mutagenesis experiments on AMPA receptors have localized structural determinants responsible for the rapid desensitization and its modulation by allosteric ligands exclusively in the S1 and S2 segments (27)(28)(29). Our present results do not exclude the possibility that the X domain can participate in the fine-tuning of channel responses, as shown by the minor difference in desensitization rates observed between the ⌬402 and full-length GluR-D constructs. However, the presence of rapid desensitization in the ⌬402 channels together with removal of desensitization by cyclothiazide of show that the N-terminal domain is not an obligatory participant in these processes. This is also consistent with a recent model for AMPA receptor desensitization, inspired and supported by high resolution structures of S1S2-ligand complexes, proposing that destabilization of a dimer interface between agonist-binding domains of neighboring subunits may underlie desensitization (23,30,31).
Assembly-Several studies have suggested a role for the X domain in the assembly of multimeric channels (10,12,32,33). Furthermore, experiments performed with separately expressed extracellular domains of GluR-D subunit indicate that the X domain is able to form dimers, consistent with a role in the assembly (11). A recent study suggests that the X domain participates in early dimerization of subunits, whereas the transmembrane part and the S2 segment are important in the formation of the functional tetramer assembly (12). Our current findings show that functional homomoric AMPA receptors can assemble from subunits lacking the entire X domain. Furthermore, our data actually suggest that homomeric ⌬402 receptors assemble at least as efficiently as wild-type GluR-D and thus argue against an assembly-promoting role for the X domain. In insect cells, the expression level of ⌬402 was consistently higher as measured by the number of high affinity AMPA binding sites. In HEK 293 cells, transfected in parallel for expression of either ⌬402 or full-length GluR-D channels, the currents recorded from outside-out patches were at the same level, even though immunoblots indicated that, in these cells, ⌬402 polypeptide is expressed at a lower level than the full-length GluR-D subunit. It should be noted, however, that most native AMPA receptors are heteromeric assemblies, and while the assembly of homomeric GluR-D channels seems to occur apparently normally in the absence of the X domain, the assembly of heteromeric subunit combinations may well require participation of the N-terminal region. Interestingly, a recent elegant study provided evidence that the assembly of homomeric AMPA receptors is a stochastic process, whereas heteromeric channels assemble with a preferred subunit stoichiometry (33). Considering the highly conserved ligand-binding domain and transmembrane segments of AMPA receptor subunits, it will be of considerable interest to test whether the more variable N-terminal domain (see below) would be able to guide the preferential formation of subunit pairs in the heteromeric assembly process.
Ligand Binding-The N-terminal domain is similar to bacterial amino acid-binding proteins and the N-terminal domain of mGluR and GABA B receptor subunits, which contain the neurotransmitter binding site (34,35). Still, previous studies on chimeric receptors (5) and soluble extracellular domains of AMPA receptor subunits (6,7,36) clearly exclude a direct participation of the N-terminal domain in agonist binding. Our previous analysis of soluble extracellular GluR-D constructs showed no differences in agonist binding affinities or kinetics between XS1S2 and S1S2 ligand-binding domains. In the intact receptor, however, structural constraints posed by domain and subunit interactions could be quite different, but the present results exclude any significant "allosteric" modulation of the ligand-binding domain by the X domain also in the full-length membrane-bound receptor.
Synaptic Interactions-Only a few extracellular synaptic interactions of iGluRs have been described, perhaps due to the inapplicability of standard yeast two-hybrid technology to extracellular proteins. The recently described interaction between the NR1 subunit of the NMDA receptor and EphB receptor tyrosine kinase (37), and the association of AMPA receptor subunits with Narp, a synaptic protein of the pentraxin family (18), have been reported to take place extracellularly, although the exact interaction sites have not been mapped within the iGluR ectodomains. Moreover, functional interactions of AMPA receptors with heparan sulfate containing proteoglycans present in the extracellular matrix have been suggested (38). As a large protein domain, which is likely to protrude further away from the membrane than the ligandbinding domain, the X domain would be in an ideal position to participate in the aforementioned and other protein-protein interactions. Furthermore, the lower level of sequence similarity between the subunits within any iGluR subclass in the X domain (ϳ55-60% sequence identity for AMPA receptor subunits) than in the S1S2 domains (Ͼ 90%) could provide room for subunit-specific functions of the X domain. The present results do not address this issue, but by excluding an essential role in the core functions of the ligand-gated channel, our results highlight the possibility of specific protein interactions as one of the functions of the X domain.
Conclusion-We conclude that the N-terminal X domain does not play a critical role in the formation of active glutamate-gated channels in homomeric GluR-D AMPA receptors. Furthermore, the ligand binding, allosteric transitions involved in the rapid activation and desensitization of AMPA receptor channel responses, as well as its regulation by cyclothiazide, can take place in the complete absence of the X domain.