Functional Assembly of AMPA and Kainate Receptors Is Mediated by Several Discrete Protein-Protein Interactions

Neuron, Vol. 31, 103–113, July 19, 2001, Copyright 2001 by Cell Press

Functional Assembly of AMPA and Kainate Receptors Is Mediated by Several Discrete Protein-Protein Interactions Gai Ayalon and Yael Stern-Bach1 Department of Anatomy and Cell Biology The Institute of Dental Sciences The Hebrew University-Hadassah School of Dental Medicine Jerusalem 91120 Israel

Summary Functional heterogeneity of ionotropic glutamate receptors arises not only from the existence of many subunits and isoforms, but also from combinatorial assembly creating channels with distinct properties. This heteromerization is subtype restricted and thought to be determined exclusively by the proximal extracellular N-terminal domain of the subunits. However, using functional assays for heteromer formation, we show that, besides the N-terminal domain, the membrane sector and the C-terminal part of S2 are critical determinants for the formation of functional channels. Our results are compatible with a model where the N-terminal domain only mediates the initial subunit associations into dimers, whereas for the assembly of the full functional tetramer, compatibility of the other regions is required. Introduction Ionotropic glutamate receptors (iGluR) mediate most of the excitatory synaptic transmission in the brain, where they also play key roles in synaptic plasticity and pathology. Based on pharmacological and electrophysiological criteria, iGluR have been traditionally classified into three major subtypes: AMPA, kainate, and NMDA receptors, named after their most selective agonist (Watkins and Evans, 1981). Molecular cloning has further identified many genes as well as RNA splice and edited variants, coding members of this family. Of these, GluR1-4 (or GluRA-D) makes up the AMPA receptors, GluR5-7 and KA1-2 comprise the kainate receptors, and NR1 together with NR2A-D form the NMDA receptors (see Hollmann and Heinemann, 1994; Dingledine et al., 1999). Expression studies in various heterologous systems demonstrated that iGluR operate as oligomeric complexes, probably tetramers (Laube et al., 1998; Mano and Teichberg, 1998; Rosenmund et al., 1998; but see Ferrer-Montiel and Montal, 1996; Premkumar and Auerbach, 1997 for possible pentameric configuration). The single subunit protein of all iGluR (Figure 1A, illustration) contains four hydrophobic domains of which M1, M3, and M4 transverse the membrane, while M2 faces the cytoplasm as a reentered loop that forms part of the channel pore (Hughes, 1994). The N terminus is extracellular and can be divided into two domains: the first ⵑ400 amino acids (NTD), which may fold similarly to the bacterial periplasmic leucine-isoleucine-valine-binding protein (LIVBP; O’Hara et al, 1993), and the later ⵑ150 amino 1

Correspondence: [email protected]

acids preceding M1 (S1), which together with the region between M3 and M4 (S2), form the extracellular glutamate binding domain (Stern-Bach et al., 1994; Kuusinen et al., 1995; Armstrong et al., 1998; and reviewed by Pass, 1998). The C-terminal part of S2 is not directly involved in agonist binding, and in AMPA receptors, due to RNA splicing, it appears in two forms, flip/flop, that mostly control receptor desensitization (Sommer et al., 1990; Mosbacher et al., 1994). Finally, the C terminus is intracellular and is involved in receptor anchoring and signal transduction (Sheng and Pak, 2000). It is well established that in vivo iGluR are predominantly or exclusively composed of heteromeric channels, a phenomenon that further increases the heterogeneity of this family. However, although the iGluR subunits share substantial sequence homology and various subunit combinations can be found within the same cell, heteromeric assemblies are confined (Wenthold et al., 1992; Brose et al., 1994; Puchalski et al., 1994)—a feature that is practically responsible for dividing the family into the three defined subtypes. What is the mechanism underlying this subtype-specific assembly? Since the restricted pattern could be also demonstrated in various non-neuronal cell types (e.g., Partin et al., 1993; Puchalski et al., 1994), it is therefore likely that an intrinsic property coded in the subunit protein causes the apparent selectivity of the assembly process. Recently, using coimmunoprecipitation assays, authors have proposed that in AMPA receptors subtypespecific heteromerization is exclusively determined by the NTD (Leuschner and Hoch, 1999). However, since the functionality of the complexes formed has not been addressed, it is possible that other intrinsic signals may also direct receptor assembly. In addition, nothing is known about the assembly mechanism itself. Here, using coexpression of chimeras with wild-type AMPA and kainate receptor subunits, we show that functional heteromers between two subunits can be formed only if, besides the NTD, other regions in the C-terminal half of the subunits also originate from the same subtype. These include the membrane sector (M1–M3 and M4) and the C-terminal part of S2, but not the agonist binding core domains (S1 and the N-terminal part of S2) or the intracellular C-terminal tail. Compatibility of either the NTDs or of the C-terminal regions alone is not sufficient for the formation of functional heteromers, although each of these regions mediate subtype-specific proteinprotein interactions. Our results are compatible with a model in which the NTD apparently mediates the initial association of the subunits into dimers, while subsequent compatibility of the other regions mediates the assembly of the full functional tetramer. Such a multistep process may provide the means for regulation and quality control of the assembly process and thereby increase its specificity. Results Functional Analysis of Heteromeric Formation between Chimeras and Wild-Type AMPA and Kainate Receptor Subunits In order to identify regions that mediate subtype-specific heteromerization, we tested the ability of chimeras

Neuron 104

Figure 1. Determination of Current-Voltage Relationship and Pharmacological Properties of R3/R6 Chimeras Coexpressed with Wild-Type AMPA and Kainate Receptor Subunits The molecular composition of the Q form subunits is shown on the left (A). Black bars correspond to GluR3 sequences and white bars to

Assembly of AMPA and Kainate Receptors 105

made of reciprocal exchanges between the AMPA receptor subunit, GluR3, and the kainate receptor subunit, GluR6, to form functional heteromers with representative wild-type subunits of each principle parent subtype. These R3/R6 chimeras form functional homomers in Xenopus oocytes as well as in HEK293 cultured cells and were previously used to localize the glutamate binding domain (Stern-Bach et al., 1994) and to identify key residues affecting receptor desensitization (Stern-Bach et al, 1998). Since AMPA and kainate receptor subunits, despite substantial sequence homology, do not coassemble even when heterologously coexpressed, the set of functional R3/R6 chimeras makes a perfect tool to identify protein regions responsible for this restricted assembly. Using functional criteria for heteromeric formation, we were able to distinguish responses mediated by heteromeric receptors from those exhibited by the homomeric counterparts and in a way that can similarly apply to both AMPA and kainate receptors. Therefore, we looked mainly at the current-voltage relationship (I-V), whose shape, in both AMPA and kainate receptors, is largely determined by a single amino acid located in M2 and which, as a result of RNA editing, can be either a glutamine (Q) or an arginine (R) (reviewed by Seeburg, 1996). The unedited Q forms exhibit a strong inward current rectification due to a voltage-dependent block by intracellular polyamines (Bowie and Mayer, 1995; Donevan and Rogawski, 1995; Kamboj et al., 1995; Koh et al., 1995), while the edited R forms are not sensitive and exhibit a linear I-V relationship or slightly outward rectification, even in various heteromeric configurations with the Q form (e.g., Hume et al., 1991; Egebjerg and Heinemann, 1993; Washburn et al., 1997). When expressed in oocytes, both GluR3Q (flip isoform) and GluR6Q homomers produce large currents upon attenuating their fast desensitization with cyclothiazide and concanavaline A, respectively (1–3 ␮A at ⫺70 mV to 1 mM glutamate) (e.g., Partin et al., 1993; Yamada and Tang, 1993), and exhibit, as expected, a strong inward current rectification (Figures 1B and 1D). In contrast, at the same conditions, the activity of GluR2R (flop isoform) was hardly detected (⬍10 nA), although its cell surface protein expression in the oocyte is similar to that of GluR3Q (see Figure 2C, lower panels). This apparent lack of macroscopic currents mainly results from the low single-channel conductance of the R form (Swanson et al., 1997) combined with the weak sensitivity of the flop isoform to cyclothiazide (Partin et al., 1994). However, as previously shown (Hume et al., 1991; Washburn et

al., 1997), when GluR2R was expressed at a 1:1 ratio with GluR3Q, we observed large macroscopic currents with a linear I-V relationship (Figure 1B, upper left, plus sign), which must reflect the predominant heteromeric formation between these two subunits. Conversely, no change in current rectification was observed when GluR2R was coexpressed with GluR6Q (Figure 1B, upper right), as expected from the lack of interaction between AMPA and kainate receptor subunits. GluR6R homomers, unlike GluR2R, produce measurable whole-cell current amplitude upon treatment with concanavaline A (50% compared to GluR6Q, data not shown; see Egebjerg and Heinemann, 1993)—a feature that may complicate the distinction between the activity mediated by the respective R/Q heteromers and that resulting from a combined activity of two separate R and Q homomers. To avoid this problem, we generated a mutant of GluR6R in which the conserved arginine (R523) in the binding pocket (Uchino et al., 1992; Armstrong et al., 1998) was replaced by lysine (K). This point mutation reduced the apparent affinity to glutamate by more than 50-fold (data not shown); therefore, with 1 mM glutamate (a saturated concentration for both GluR3Q and GluR6Q homomers at the conditions used), no currents could be detected for homomeric R6(R523K)R. However, when it was coexpressed with GluR6Q, we obtained a linear I-V relationship (Figure 1D, top right), and as expected from the subtype-restricted assembly, R6(R523K)R did not attenuate the current rectification of GluR3Q (Figure 1D, top left). Having set the control conditions, we next tested the ability of each of these practically “inactive” or “silent” GluR2R and GluR6(R523K)R subunits to abolish the inward current rectification of four R3/R6 Q form chimeras (Figure 1A). Each of these chimeras forms functional homomeric receptors with expression levels comparable to those of the wild-type GluR3Q and GluR6Q homomers. Reciprocally they divide the subunit polypeptide into three regions: NTD, S1, and the C-terminal half that includes the membrane domains, the extracellular S2, and the intracellular C-terminal tail. As shown in Figure 1B, no significant change in the I-V relationships was observed when GluR2R was coexpressed with R6NTDR3Q, a chimera in which the NTD originates from GluR6 and the rest of the polypeptide from GluR3 (Figure 1B, middle left), or when coexpressed with the reciprocal chimera, R3NTDR6Q, in which the NTD is the only domain compatible to GluR2R (Figure 1B, middle right). In contrast, GluR2R abolished the rectification of R3(R6S1)Q, a chimera in which only S1 originated from GluR6 (Figure 1B,

GluR6. NTD represents the first ⵑ400 amino acids, S1 and S2 the ligand binding domains; the long vertical bars are the three transmembrane domains M1, M3, and M4, and the short vertical bar corresponds to the reentered channel loop, M2. The numbers above each chimera indicate the residues forming the respective junction according to the color code. All chimeras contained the flip module (where applicable) while GluR2R and GluR3R the flop. GluR6*R corresponds to GluR6(R523K)R. The table on the right summarizes the data shown in (B)–(E). A linear I-V relationship (B–D) upon coexpression with a specific R form subunit and a positive response to AMPA (100 ␮M) upon coexpression with KA2 (E) is indicated by “Yes,” while no change in phenotype relative to the homomeric counterparts is represented by “No.” n.d.1, not determined. n.d.2, not determined because the chimeric subunit forms homomers sensitive to AMPA. Subunits were injected alone (⫺) or at 1:1 ratio (⫹) with the respective subunits, as indicated. Recordings were made as described in Experimental Procedures. The protein ratio was verified by Western blot analysis on whole-cell protein fraction (data not shown, but see Figure 2C). Each I-V curve is an average of three oocytes injected with tagged subunits (HSV-Q form versus FLAG-R form). Similar results were obtained with untagged subunits, and the total number of experiments is given for each pair (N; bottom right). Responses to 1 mM glutamate of the Q form subunits were 1000–5000 nA (at ⫺70 mV).

Neuron 106

bottom right). Coexpression of the chimeras with GluR3R produced similar results to that obtained with GluR2R (Figure 1C), thus excluding interference from the sequence diversity within the AMPA receptor subtype. For the mirror image experiments of coexpressing the chimeras with the kainate receptor subunit R6(R523K)R (Figure 1D), we obtained, in principle, similar results to those observed with the AMPA receptor subunits GluR2R and GluR3R. That is, when coexpressed, R6(R523K)R could not abolish the inward current rectification of any of the four chimeras, although its presence in each set of oocytes was verified by Western blot analysis (data not shown). The chimera R6(R3S1), which we would predict to coassemble with GluR6, does not form functional homomers (Stern-Bach et al., 1994), and therefore was omitted from the assay. Finally, we also tested possible assembly with the kainate receptor subunit KA2. This subunit does not yield functional homomeric channels in oocytes (nor in other heterologous systems), but when coexpressed with GluR6, it enables responses to AMPA, a feature not exhibited by homomeric GluR6 receptors (e.g., Herb et al., 1992; Figure 1E, left). Of the chimeras shown in Figure 1A, only R3NTDR6 forms homomers that are not activated by AMPA, and its coexpression with KA2 fails to convert this phenotype (Figure 1E, right), even though the entire agonist binding site of this chimera is that of GluR6. Taken together, the analysis of the chimeras shown in Figure 1 identified the NTD as well as the region C-terminal to S1, but not S1, as major determinants for subtype-specific assembly and demonstrated that compatibility in only one of these regions is not sufficient to obtain functional heteromers. Functional and Biochemical Characterization of Subtype-Specific Intersubunit Interaction of Fully and Partially Compatible Subunits The apparent lack of functional heteromerization when either the NTD or the region C-terminal to S1 is the only compatible domain can result from a complete lack of interaction between the two subunits or from partial interactions that are not sufficient to generate functional channels. Because the activity of the R form homomers cannot be detected in our experimental conditions, in the case of no interaction, the amplitude of whole-cell current will be determined only by the amount of the Q form subunit in each configuration. In contrast, in case of insufficient interactions, we would expect a relative reduction in whole-cell currents as a result of recruitment of the functional Q form subunits into truly inactive complexes with the R form. To examine this issue in detail, we coexpressed the Q form chimeras with the “silent” GluR2R subunit at various ratios by injecting a fixed amount of the Q form subunit mixed with increasing amounts of GluR2R and measured whole-cell current amplitudes as well as the degree of current rectification. At these expression conditions, since the total amount of the cRNA injected was kept below saturation, each set of oocytes expressed the same amount of the Q form protein (Figure 2C). Therefore, the current amplitude obtained for each ratio (I) was divided by the value obtained for oocytes injected with the Q form alone (I0).

For the control experiment of coexpressing GluR6Q with GluR2R, there was no significant change in I/I0 even at six times excess of GluR2R (Figure 2A) as well as no change in current rectification (Figure 2B), as expected from the lack of interaction between wild-type AMPA and kainate receptor subunits. Similar results were obtained for the coexpression of R6NTDR3Q or R6TM1R3Q with GluR2R (data not shown), suggesting that these two chimeras, in which the NTD originates from GluR6, do not interact with GluR2R, and that is why current rectification was not attenuated (Figure 1B). On the other hand, for the coexpression of R3NTDR6Q and GluR2R, subunits that are compatible only in the NTD, we observed a gradual decrease in I/I0 as the amount of GluR2R increased (Figure 2A), with almost no effect on current rectification (Figure 2B). Therefore, the compatible NTDs interact with each other, but these interactions alone are insufficient and result in nonfunctional complexes. Only when the amount of GluR2R was in three times excess was a small change in the rectification observed (p ⬍ 0.01), which may indicate the formation of a small population of functional complexes. However, since the remaining macroscopic currents were relatively small (⬍100 nA), part of this phenotype may be contributed by the large population of homomeric GluR2R. The reduction in activity upon coexpression is not a result of a lower surface expression (Figure 2C). In fact, we observed a higher signal for the coexpressed subunits compared to each of the homomeric expressions. Currently, without further analysis, the reason for this phenomenon is unclear. Nevertheless, it indicates that even though these nonfunctional complexes are delivered to the oocyte cell surface, they are treated differently from functional channels. In order to gain some insights into the type of interactions mediated by the NTD, we calculated the theoretical reduction in I/I0 expected for various receptor stochiometries assuming a binomial distribution (see details in Experimental Procedures). Interestingly, the observed reduction in I/I0 was much lower than what would be expected if the compatible NTDs assemble into the correct oligomeric order being either a tetramer or a pentamer (Figure 2A, traces 4 and 5). On the other hand, the data fit much better with a theoretical formation of dimers (trace 2). At high amounts of GluR2R, however, the deviation of the data from trace 2 may indicate limited formation of complexes of higher order. In contrast to R3NTDR6Q, for the chimera R3(R6S1)Q, in which both the NTD and the C-terminal half of the subunit are compatible to GluR2R, coexpression at similar conditions showed a gradual increase in I/I0 with increasing amounts of GluR2R (Figure 2A). In parallel, current rectification was significantly attenuated, even when small amounts of GluR2R were added (Figure 2B), indicating predominant formation of functional heteromers at all ratios. Because the low-single channel conductance of GluR2R is dominated by the high conductance of the Q form subunit when forming heteromers (Swanson et al., 1997; Washburn et al., 1997), the gradual increase in the macroscopic current is a result of the formation of more high-conducting heteromeric channels as the amount of GluR2R was increased. This increase in activity correlates with formation of tetramers (Figure 2A, trace 4), a finding that is in line with

Assembly of AMPA and Kainate Receptors 107

Figure 2. Coexpression of GluR2 with GluR6, R3NTDR6, or R3(R6S1) at Various Protein Ratios (A) Whole-cell current measurements on oocytes injected with a fixed amount of cRNA coding for the respective Q form subunit (2–3 ng), together with water or different amounts of GluR2R cRNA (in an equivalent volume). Recordings were performed as described in Experimental Procedures. I/I0 defines the value measured at ⫺70 mV for a given subunit ratio (I), divided by the value obtained from oocytes injected with the Q form subunit alone (I0). Data for each point is an average of 25–53 oocytes from 3–7 batches. Similar results were obtained when the total cRNA injected per oocyte was kept constant by supplementing with cRNA coding for the NMDA receptor subunit NR1a. The dashed traces are theoretical curves generated for the formation of dimers (2), trimers (3), tetramers (4), and pentamers (5) as described in Experimental Procedures. (B) Representative I-V curves for the experiments summarized in (A). The Q:R ratio is indicated on the right of each curve. Whole-cell current amplitudes measured at ⫺70 mV were (in nA): 1940 ⫾ 240 for GluR6Q, 2384 ⫾ 283 for R3NTDR6Q, and 1354 ⫾ 219 for R3(R6S1)Q injected without GluR2R. (C) Western blot analysis on whole-cell and biotin labeled (surface) protein fractions (obtained from 10 oocytes), were performed as described in Experimental Procedures on the same oocytes shown in (B).

our current assumption on the functional core stochiometry of the receptor. The deviation of the data at high ratios may be a result of lower single-channel conductance of channels having more than one copy of GluR2R, or of the fact that assembly is not completely random. Finally, in order to directly examine the observed pro-

tein-protein interactions between the chimeras and the wild-type parent subunits, we used biochemical coprecipitation experiments (Figure 3). To unify the experimental design the “precipitating” subunits were tagged with FLAG (F) and a six-histidine tail (His) for detection and purification purposes, respectively, while the “copreci-

Neuron 108

Figure 3. Coprecipitation of Chimeras with Wild-Type AMPA and Kainate Receptor Subunits Coprecipitation experiments were done as described in Experimental Procedures. The identity of the precipitating subunit is indicated on the top, and the coprecipitating subunits are marked below. T, the total protein fraction loaded on the beads; P, the precipitated fraction. Equal amounts of T and P were separated on 8% SDS-PAGE gels, blotted and reacted with anti-HSV (␣H) or anti-FLAG (␣F) antibodies as indicated.

pitating” subunits were tagged with HSV (H). Thus, we were able to detect each subunit in the pairwise mixtures using the respective anti-tag antibodies, regardless of the subunit’s structure. Following expression in HEK293 cells, subunits lacking the His tag did not retain on the metal beads (Figure 3A), while those expressing it fully precipitated at the condition used (⬎95%; Figures 3B and 3C, lower panels). As expected, when coexpressed at a 1:1 ratio, HGluR3 fully coprecipitated with FGluR3His (95% ⫾ 5%), while no coprecipitation was observed for HGluR6 (⬍2%) (Figure 3B). At the same conditions, however, only partial coprecipitation was observed with H R3NTDR6 (54% ⫾ 7%) and with HR6TM1R3 (15% ⫾ 3%). Similar results were obtained for coprecipitation with FR6TM1R3His (Figure 3C). That is, for the fully compatible HR6TM1R3, we observed 95% ⫾ 3% coprecipitation, while for the partially compatible HGluR6 and H GluR3 we obtained 52% ⫾ 4% and 12% ⫾ 3% coprecipitation, respectively. Similar results were obtained for coprecipitation of the chimeras with GluR1 (N. Brose, Y.S.-B., and S.F. Heinemann, unpublished data). The partial 52%–55% coprecipitation of subunits that are compatible only in their NTD correlates with our electrophysiological measurements in which, at the same expression ratio (1:1), we observed ⵑ58% inhibition of whole-cell currents (Figure 2A). The weak interactions observed for the region C-terminal to S1 when being the only domain compatible (12%–15% coprecipitation) were not detected by the electrophysiological measurements, a discrepancy that may be due to the different expression system used in each case. In summary, the results shown in Figures 2 and 3 indicate that compatible NTDs mediate subtype-specific protein-protein interactions, but these interactions alone are not sufficient to produce functional heteromers and apparently result in partially assembled complexes, probably dimers. For the region C-terminal to S1, in the absence of compatibility between the NTDs, we observed only weak intersubunit interactions. However, since it is clear that without compatibility in this region functional heteromers are not formed, it is possible that we cannot detect these interactions because they are dependent on conditions created by the NTD interactions. Characterization of Determinants C-Terminal to S1 Important for Functional Assembly of Homomeric and Heteromeric Channels In order to identify the molecular determinants in the region C-terminal to S1 that are responsible for subtype

specific heteromerization, we generated several chimeras dividing this region into four distinct subdomains (Figure 4): C1, the region containing the transmembrane domains M1 and M3 and the channel forming pore M2 (M1–M3); C2, the region between M3 and M4 (S2) that forms part of the agonist binding domain; C3, the transmembrane domain M4; and C4, the intracellular C-terminal tail. Because the identity of the NTD is critical for heteromeric assembly, these C1–C4 alternations were made in R3(R6S1)Q (Figure 4A) and in GluR6Q (Figure 4B), respectively. This created two sets of reciprocal mutants that enabled us to test for heteromerization with AMPA receptor subunits, on one hand, and with kainate receptor subunits, on the other. Using the tests described in Figure 1, we found that alternations of C1, C2, and C3 prevented the formation of functional heteromers. In contrast, functional heteromer formation was unaffected by an alternation confined to C4 (see chimera R3(R6S1)-C4, Figure 4A), a result that excludes the involvement of the intracellular C-terminal tail in subtype specific assembly. Moreover, by splitting C2 in half (see chimera R6-C2N, Figure 4B), we found that the N-terminal part of this region (forming the smaller lobe of the agonist binding domain) is also not involved in this process. The complementary chimera was nonfunctional (R6-C2C; data not shown). Interestingly, we found that chimeras having C1 (M1–M3) or C3 (M4) of the opposite subtype had 5- to 10-fold lower homomeric activity compared to subunits having both domains of the same subtype (data not shown). Since the level and kinetics of protein synthesis as well as affinity to glutamate were not altered in these chimeras (data not shown), it indicates that intrasubunit compatibility between the membrane domains is required for full homomeric activity. Finally, in order to further dissect the M1–M3 region (C1), we generated two more mutants of R3(R6S1)Q in which either M1 or M2 (including its two connecting loops) was replaced by the GluR6 sequence. M3, except for the conserved change of valine to isoleucine, is identical between GluR3 and GluR6 and was therefore disregarded. Of the two mutants, the M1 exchange resulted in a nonfunctional mutant while the M2 exchange mutant was only partially active (5%, compare to R3(R6S1)Q) and also did not functionally assemble with GluR2R (data not shown). Therefore, more refined site-directed mutagenesis is needed to resolve the contribution of specific residues. In summary, the dissection of the subunit’s C-terminal half disclosed the membrane regions M1–M3 and M4

Assembly of AMPA and Kainate Receptors 109

Figure 4. Refinement of C-Terminal Determinants Important for Functional Heteromerization The region C-terminal to M1 has been divided into four domains as marked: C1, M1–M3; C2, S2; C3, M4; and C4, the intracellular C terminus. C1–C4 alternations were made in R3(R6S1)Q (A) and GluR6Q (B), respectively. All mutants contained the flip module where applicable. Chimeras were expressed at 1:1 ratio with either GluR2R, GluR6(R523K)R, or KA2, as indicated. Yes/No indicate whether functional heteromers were formed between a specific pair based on the I-V relationships or sensitivity to AMPA as described in Figure 1 (n ⫽ 5–10, N ⫽ 2–3). n.d.1, not determined because the chimera was nonfunctional as homomer. n.d.2, not determined because the chimera forms homomers sensitive to AMPA.

as well as the C-terminal half of S2 as playing critical roles in receptor assembly, while excluding the involvement of the agonist binding region of S2 and the intracellular C-terminal tail. Discussion In this study we have identified multiple domains throughout the subunit polypeptide that play key roles in the assembly of AMPA and kainate receptors. This has been achieved by testing the ability of several chimeras of the AMPA receptor subunit GluR3 and the kainate receptor subunit GluR6 to form functional channels with each one of the parent wild-type subunits as well as with the homologous subunits GluR2 and KA2, respectively. Therefore, the results of this study are likely to apply to the assembly of the other AMPA and kainate receptor subunits as well. By looking at chimeras that split the subunit into two or three broad regions, our first finding was that functional heteromers between a chimera and a wild-type subunit can be obtained only if their NTD and the region C-terminal to S1 both originate from the same subtype (Figure 1, see summary in [A]). This finding suggests that the heteromerization process is complex and not just determined by the NTD as previously proposed (Leuschner and Hoch 1999). What is the reason for the different conclusions reached in our study? Leuschner and Hoch (1999) based their conclusion on the finding that chimeras and truncated subunits that are compatible only in the NTD coprecipitated to a similar extent as fully compatible subunits. However, this assay does not distinguish between functional and nonfunctional interactions. In addition, in their study, the precipitating subunit was expressed in excess to the other subunit. This together with the observed

high nonspecific background precipitation may mask small differences in the extent of coprecipitation. In our study, in agreement with Leuschner and Hoch, we found that subunits compatible only in the NTD interact with each other (Figure 3), but the complexes formed are nonfunctional. This is evident from the significant reduction in whole-cell current upon coexpression (Figure 2A, middle), with no evidence for heteromeric activity (Figure 2B, middle). Only when there is compatibility in the C-terminal half of the subunits as well are functional heteromers formed (Figures 2A and 2B, right). Furthermore, in our coprecipitation experiments (Figure 3), at a 1:1 ratio we found only partial coprecipitation of subunits that are compatible only in the NTDs as compared to fully compatible subunits (55% versus 95%). This result indicates once more that when downstream compatibility is absent, the interaction between the subunits is not optimal. One of the most striking observations of our study is that the reduction in whole-cell current observed, when only the NTDs are compatible is lower than what would be expected if the nonfunctional complexes formed between the two subunits, would consist of tetramers or pentamers (Figure 2A, middle, traces 4 and 5). The data, on the other hand, fits better with a theoretical formation of dimers (trace 2). Conversely, when both domains are compatible and, thus, functional heteromeric channels are formed (Figure 2B, right), the titration of whole-cell current fits well with the formation of tetramers (Figure 2A, right, trace 4). The most straightforward explanation of these results is (1) tetramers are the functional unit in assembled AMPA and kainate receptors—a finding that is in line with other studies on receptor stochiometry (Laube et al., 1998; Mano and Teichberg, 1998; Rosenmund et al., 1998)—and (2) compatible NTDs (with no

Neuron 110

downstream compatibility) are only able to form dimers. At this point we can not rule out the alternative that the “dimer-like” behavior—observed in Figure 2A, middle—is due to weaker or less stable intersubunit interactions when the NTD is the only compatible domain. This would lead to a smaller fraction of the complexes being heterotetramers and the rest pure homotetramers. This issue may be eventually resolved by biochemical isolation and characterization of intermediate complexes. Support for our interpretation that the NTD can assemble only as dimers is provided by two recent biochemical studies showing that truncated NTD of GluR1 or GluR4, alone or linked to S1-S2, predominantly forms homodimers in solution (Kuusinen et al., 1999; Wells et al., 2001). In addition, recent crystallization of the N-terminal domain of the metabotropic glutamate receptor 1, which is homologous to the NTD of the iGluRs (O’Hara et al., 1993), suggests that this domain exists as dimers (Kunishima et al., 2000). In contrast to the NTD interactions, in our coprecipitation assays, we observed only weak intersubunit interactions mediated by the C-terminal region, when compatibility between the NTDs is absent (Figure 3). However, since it is clear that without compatibility in this region functional channels are not formed, this apparent weak interaction may indicate that, during the assembly of native channels, the interactions of the NTDs are a prerequisite needed for the assembly of the C-terminal half. This hypothesis is in line with the fact that chimeric subunits, having the NTD from one subtype and the C-terminal half from the other, do form fully functional homomeric channels as the wild-type parent subunits (Figure 1; Stern-Bach et al., 1994). This indicates that compatibility is required within each of these domains but not in-between. By testing chimeras having limited substitutions in the C-terminal half of the subunit, we found that compatibility in the membrane domains (M1–M3 and M4), as well as in the C-terminal half of S2 (containing the flip/flop module in AMPA receptors), is important for functional heteromeric assembly (Figure 4). This excludes the N-terminal half of S2 that is directly involved in agonist binding (Stern-Bach et al., 1994; Armstrong et al., 1998) and the intracellular C-terminal tail that binds to scaffold and other regulatory proteins (Sheng and Pak, 2000). Thus, even though multiple domains are required for assembly, the agonist binding domain and the carboxyl terminus are not among them. Our finding of the involvement of the distal part of S2 in intersubunit interactions is in harmony with a recent structural analysis of dimeric crystals of the miniprotein S1-S2 of GluR2, showing that this region forms the dimer interface (Armstrong and Gouaux, 2000). Initial attempts to determine the contribution of specific regions in M1–M3 for subunit assembly indicated that the identity of both M1 and M2 is important (data not shown). Therefore, the role of M1 is more than just providing a membrane anchor that permits or facilitates the oligomerization of the NTD, as suggested previously (Leuschner and Hoch, 1999). Taken together, our results are compatible with a model (Figure 5), in which the assembly of AMPA and kainate receptors proceeds in at least two basic sequential steps: step 1, subunit dimerization that is mediated by the NTD; step 2, dimer:dimer association that is mediated by the C-terminal determinants and stabilized by

secondary NTD interactions. The model illustrates homomeric assembly of representative subunits (Figures 5A and 5B), which can similarly apply to heteromeric assemblies within each of the subtypes. When AMPA and kainate receptor subunits are expressed in the same cell (Figure 5C), noncompatibility in the NTDs primarily prevents heterodimer association, resulting in the formation of homodimers only. The two distinct homodimers can then assemble only with themselves, but not with each other because of the incompatible C termini. In agreement with our experiments using chimeras (Figure 1), due to this sequence of events, subunits having the NTD of one subtype and the C-terminal determinants of the other can form fully functional homomers (Figures 5D and 5E), but not functional heteromers with wildtype subunits (Figures 5F and 5G). A chimera and a wildtype subunit compatible in the C termini but not in the NTDs (Figure 5F) can not heterodimerize, whereas each one can form stable homodimers. The homodimers are then able, in principle, to interact through the compatible C termini; however, in the absence of secondary compatibility in the NTDs, these complexes are not stable and thus dissociate. Evidence for such nonstable heterotetramers is provided by the residual coprecipitation of R6TM1R3 with GluR3 and vice versa (15%; Figures 3B and 3C). On the other hand, a chimera and a wildtype subunit compatible only in the NTDs (Figure 5G), can form stable heterodimers as well as the two distinct homodimers, but the subsequent dimer:dimer associations can take place only between dimers that have fully compatible C termini. This results in the formation of two separate populations of functional homotetramers and a population of nonfunctional heterodimers, in agreement with the data obtained for the coexpression of R3NTDR6Q with GluR2R (Figure 2A, middle). In vivo, having multistep signals spread throughout the subunit may provide the means for regulation and quality control of the assembly process and thereby increase its specificity. In addition, while the NTD interactions may serve as an initial rapid screen between the possible pairs, it is possible that certain pairs may be preferred or more stable than others, and therefore, the final tetrameric arrangement will depend on the additional downstream signals. This, in turn, may further restrict the possible configuration of native channels within each of the subtypes. Such restrictions have in fact been observed. In the CA1/CA2 pyramidal neurons, it was found that AMPA receptors are mainly composed of GluR1 and GluR2 or GluR2 and GluR3, while very few complexes contained both GluR1 and GluR3 or only GluR1 (Wenthold et al., 1996). Furthermore, heteromers of GluR1 and GluR2 seem to preferentially assemble into symmetric tetramers containing two copies of each subunit (Mansour et al., submitted). In addition, in the NMDA receptors, which conduct only as heteromers made of NR1 with any of NR2A-D subunits, two NR1 and two NR2 subunits make up the functional complex (Behe et al., 1995; Laube et al., 1998). Preferential dimer:dimer arrangement is also proposed for the tetrameric cyclic nucleotide-gated channels (Liu et al., 1998) and the voltage-gated potassium channels (Tu and Deutsch, 1999). However, in the latter case the tetramerization signals are contained in the N-terminal domain only (reviewed in Papazian, 1999). For the

Assembly of AMPA and Kainate Receptors 111

Figure 5. Schematic Model for Native and Chimeric Receptor Assembly The model suggests two basic steps in receptor assembly: step 1, subunit dimerization that is mediated by the NTD; step 2, dimer: dimer association that is mediated by the C-terminal determinants and stabilized by secondary NTD interactions. (A)–(G) illustrates the different homomeric (A, B, D, and E) and heteromeric (C, F, and G) expressions of wild-type and chimeric subunits examined in this study. Black and white shapes represent AMPA and kainate receptor subunits, respectively. The intermediate dimers (step 1) and the resulting tetramers (step 2) are shown for each case. Unfavored or unstable complexes are crossed. The probability of occurrence of the other species is undetermined. See Discussion for more details.

pentameric nicotinic acetylcholine receptors heterodimer formation—mediated by the N terminus—is proposed to be an intermediate step (Green, 1999; Keller and Taylor, 1999). Similar to our findings, it has been recently shown that without proper signals in the C-terminal half of the ␥ subunit of the nicotininc acetylcholine receptors, the intermediate dimers do not undergo further assembly (Eertmoed and Green, 1999). Taken together, it appears that although the assembly of ion channels may be similar in some ways (for example, almost all involve initial associations of the N-terminal domain), the specific steps of the process are unique for each family. We anticipate that future studies determining the role of subdomains of both the NTD and the C-terminal region will lead to further refinement of our proposed model for iGluR assembly. Experimental Procedures Molecular Biology and In Vitro cRNA Transcription The chimeras shown in Figure 1 were previously described (SternBach et al., 1994, 1998), where chimeras R3NTDR6 and R6NTDR3 correspond to R3KBPR6 and R6KBPR3. The other chimeras used in this study were constructed by a similar approach. For each construct, the amino acids at the appropriate junctions are indicated in the figures and numbered relative to the first methionine in the ORF. GluR6(R523K)R was constructed according to the method developed by Stratagene (QuickChange; Stratagene, La Jolla, CA). The epitope tags HSV (QPELAPEDPED) and FLAG (DYKDDDDK) were inserted in frame at the N terminus of GluR3 after R36 and at GluR6 after P49, using PCR overlap extension, thus creating HGluR6, HGluR3, F GluR6, and FGluR3 tagged subunits. Similarly, a six-histidine tail (His) was added to FGluR3 after the last amino acid, generating F GluR3His. All mutations and insertions were confirmed by sequencing through the cassette insert. The tagged subunits were indistin-

guishable from the respective wild-types based on pharmacology measured in oocytes and by kinetic parameters measured in HEK293 cells (R. Petroski, Y. S.-B., and S.F. Heinemann, unpublished data). All constructs shown here (except for KA2) were inserted in pGEMHE (a gift from E. Liman), which was used for expression in Xenopus oocytes. Several constructs were also subcloned into pCDNA3 (InVitrogen, San Diego, CA) for expression in HEK293 cells. FGluR2R-flop and GluR3R-flop were obtained form the laboratory of S. Heinemann. The original R3(Q612R) (Hume et al., 1991) was subcloned from SK⫺ into pGEMHE using NsiI-SalI digest of F GluR3. Plasmids were linearized with NheI and capped cRNA was transcribed in vitro with T7 RNA polymerase (mMessage mMachine; Ambion, Austin, TX). The yield and quality of transcripts were assessed by agarose gel electrophoresis and ethidium bromide staining.

Oocyte Preparation and Electrophysiology Stage V–VI Xenopus laevis oocytes were prepared as previously described (Stern-Bach et al., 1994). Oocytes were injected up to 24 hr after preparation, with 1–20 ng cRNA in 50 nl/oocyte (linear range of expression), and assayed 3–5 days later. Two electrode voltageclamp recordings were carried out at room temperature, using GeneClamp500 connected to DIGIDATA1200 and pCLAMP6 (Axon Instruments, Foster City, CA). Electrodes (Sutter Instruments, Novato, CA) were filled with 3 M KCl and had resistance of ⵑ1 M⍀. Oocytes were continuously perfused with recording solution containing 10 mM HEPES, pH 7.4, 90 mM NaCl, 1 mM KCl, and 1.5 mM CaCl2 (Figures 1 and 4) or 1.8 mM MgCl2 (Figure 2; to minimize the activation of the endogenous Ca2⫹-gated Cl⫺ channels). To avoid interference from receptor desensitization, oocytes were usually treated with concanavaline A (1 mg/ml, 5–10 min before recording; Sigma, St. Louis, MO) and cyclothiazide (0.1 mM mixed in the agonist solution; RBI, Natick, MA). I-V relationships were obtained by applying 1 s voltage ramps (⫺70 mV to ⫹40 or ⫹70 mV, as indicated) on plateau responses to 1 mM glutamate. Measurements in the presence of agonist were subtracted from the average values obtained before and after agonist application, and normalized to the value

Neuron 112

obtained at ⫺70 mV. Sensitivity to AMPA (Figures 1 and 4) was tested by applying 100 ␮M AMPA (holding potential ⫺70 mV). DATA were analyzed using pCLAMP6 and ORIGIN 4.0 software (OriginLab, Northampton, MA). Cell Surface Labeling and Determination of Subunit Protein Ratios Using Western Blot Analysis Cell surface labeling of proteins with biotin, SDS-PAGE, and Western blot analysis were performed as previously described (Stern-Bach et al., 1994). Primary antibodies and their respective working dilution were: rabbit anti-rat GluR2/3, 1:2000 (C-terminal; Chemicon, Temecula, CA), rabbit anti-rat GluR6/7 1:1000 (C-terminal), and rabbit anti-rat KA2, 1:500 (Upstate Biotechnology, Lake Placid, NY), mouse anti-FLAG • M2 1:1000 (Kodak Scientific Imaging Systems, New Haven, CT), mouse anti-HSV-Tag 1:6000 (Novagen, Madison, WI). Quantification of the tag-subunit protein was relative to a control protein supplied with anti-FLAG and anti-HSV antibodies and at the dilution used both antibodies were similarly sensitive. The protein ratio between coexpressed subunits was determined by densitometric analysis using NIH Image 1.52 software on scanned autoradiograms. Transient Expression in HEK293 Cultured Cells and Coprecipitation Assays Cells were cultured in 5 ml growth media on 26 cm2 flasks, at 5% CO2, 37⬚C according to standard protocols. Transfections were done on 30%–50% confluent cultures according to the calcium-phosphate method, with 4 ␮g/flask total plasmid cDNA (midrange of linear protein expression). After 24 hr incubation, cells were washed and incubated in fresh media for additional 12–24 hr. For the coprecipitation assays, cells were washed with warm PBS, scraped, collected by centrifugation, and homogenized in 550 ␮l/flask with icecold solubilization buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1.5% Triton X-100, 0.05% SDS) in the presence of a cocktail of protease inhibitors (Cφmplete; Roche Diagnostics, GmbH, Germany). Homogenates were incubated for 30 min on ice, and solubilized proteins were separated by 20 min ultracentrifugation at 45 K, 4⬚C. Aliquots, 150–200 ␮l of the clear supernatants (T fraction), were loaded on 50 ␮l (bed volume) of prewashed Talon beads (for separation using the His tail; Clontech, Palo Alto, CA) and incubated for 45 min at 4⬚C with shaking. Beads were then washed twice with 1 ml ice-cold solubilization buffer with 20 mM Imidazol, and bound proteins were released by 2 min incubation at room temperature with 100 ␮l elution buffer (SDS-PAGE loading buffer with 1% Triton X-100 and 200 mM Imidazol) and collected after a short centrifugation (P fraction). Western blot analysis was performed as described above. The extent of precipitation of each subunit from the total amount loaded has been determined using densitometry analysis on scanned autoradiograms obtained from three independent experiments done in duplicates. Determination of the Theoretical Distribution of Homomers and Heteromers Forming Different Stochiometric Orders In the coexpression experiments shown in Figure 2, the desired Q:R ratio was achieved by adding increasing amounts of the R form subunit (nR) to a basic constant amount of the Q form subunit (nQ). Assuming an equal efficiency and random assembly, the theoretical fraction of the homomeric Q form is FQ ⫽ (nQ /nQ ⫹ nR)X and that of homomeric R form is FR ⫽ (nR /nQ ⫹ nR)X, where X is the putative oligomeric stochiometry. When expressed alone (nQ ⫽ 1, nR ⫽ 0), the I0 current is generated by nQ/y channels. However, when adding nR subunits of the R form to meet the desired ratio, the number of putative channels is (nQ ⫹ nR)/y, thus increasing by the factor ␣ ⫽ (nQ ⫹ nR)/nQ. Thereby, if only homomeric Q form subunits contribute to the macroscopic currents, I/I0 ⫽ ␣FQ ⫽ (nQ ⫹ nR)/nQ * (nQ /(nQ ⫹ nR)X. Thus, for nQ ⫽ 1, I/I0 ⫽ (1 ⫹ nR) * (1/1 ⫹ nR)X. Likewise, if all combinations but homomeric R form contribute to the macroscopic current I/I0 ⫽ ␣(1 ⫺ FR) ⫽ (nQ ⫹ nR)/nQ * (1 ⫺ (nR /(nQ ⫹ nR)X), and for nQ ⫽ 1, I/I0 ⫽ (1 ⫹ nR) * (1 ⫺ (nR /1 ⫹ nR)X). For example, for X ⫽ 4 (tetramer) when an equal amount of silent R form (nR ⫽ 1) is added to the active Q form (nQ ⫽ 1), for the first case: I/I0 ⫽ 2 * (1/2)4 ⫽ 0.125 (or 12.5%), and for the second case: I/I0 ⫽ 2 * (1 ⫺ (1/2)4) ⫽ 1.875 (or 187.5%).

Acknowledgments We thank S. Heinemann for his support at the early stages of the project and for the gifts of GluR3(Q621R) and FGluR2R, R. Petroski and N. Brose for their help with preliminary experiments, M. Geylis for excellent technical assistance, and A. Kluger for help with statistics. We further express our gratitude to B. Kanner for his help in revising the manuscript and to J. Lerma, R. Rahamimoff, Z. Siegfried, M. Treinin, and R. Wenthold for their comments. This work was supported by THE ISRAEL SCIENCE FOUNDATION (grant no. 640/97) and by the Bernard Katz Minerva Center for Cell Biophysics. Received January 19, 2001; revised May 22, 2001. References Armstrong, N., Sun, Y., Chen, G.-Q., and Gouaux, E. (1998). Structure of a glutamate-receptor ligand-binding core in complex with kainate. Nature 395, 913–917. Armstrong, N., and Gouaux, E. (2000). Mechanisms for activation and antagonism of an AMPA-sensitive glutamate receptor: crystal structure of the GluR2 ligand binding core. Neuron 28, 165–181. Behe, P., Stern, P., Wyllie, D.J.A., Nassar, M., Schoepfer, R., and Colquhoum, D. (1995). Determinantion of NMDA NR1 subunit copy number in recombinant NMDA receptors. Proc. R. Soc. Lond. B 262, 205–213. Bowie, D., and Mayer, M.L. (1995). Inward rectification of both AMPA and kainate subtype glutamate receptors generated by polyaminemediated ion channel block. Neuron 15, 453–462. Brose, N., Huntley, G.W., Stern-Bach, Y., Sharma, G., Morrison, J.H., and Heinemann, S.F. (1994). Differential assembly of coexpressed glutamate receptor subunits in neurons of rat cerebral cortex. J. Biol. Chem. 269, 16780–16784. Dingledine, R., Broges, K., Bowie, D., and Traynelis, S.F. (1999). The glutamate receptor ion channels. Pharmacol. Rev. 51, 7–61. Donevan, S.D., and Rogawski, M.A. (1995). Intracellular polyamines mediate inward rectification of Ca(2⫹)-permeable alpha-amino-3hydroxy-5-methyl-4-isoxazolepropionic acid receptors. Proc. Natl. Acad. Sci. USA. 92, 9298–9302. Eertmoed, A.L., and Green, W.N. (1999). Nicotinic receptor assembly requires multiple regions throughout the gamma subunit. J. Neurosci. 19, 6298–6308. Egebjerg, J., and Heinemann, S.F. (1993). Ca2⫹ permeability of unedited and edited versions of the kainate selective glutamate receptor GluR6. Proc. Natl. Acad. Sci. USA 90, 755–759. Ferrer-Montiel, A.V., and Montal, M. (1996). Pentameric subunit stoichiometry of a neuronal glutamate receptor. Proc. Nat. Acad. Sci. USA. 93, 2741–2714. Green, W. (1999). Ion channel assembly: creating structures that function. J. Gen. Physiol. 113, 163–169. Herb, A., Burnashev, N., Werner, P., Sakmann, B., Wisden, W., and Seeburg, P.H. (1992). The KA-2 subunit of excitatory amino acid receptors shows widespread expression in brain and forms ion channels with distantly related subunits. Neuron 8, 775–785. Hollmann, M., and Heinemann, S. (1994). Cloned glutamate receptors. Annu. Rev. Neurosci. 17, 31–108. Hughes, T.E. (1994). Transmembrane topology of the glutamate receptors. A tale of novel twists and turns. J. Mol. Neurosci. 95, 211–217. Hume, R.I., Dingledine, R., and Heinemann, S.F. (1991). Identification of a site in glutamate receptor subunits that control calcium permeability. Science 253, 1028–1031. Kamboj, S.K., Swanson, G.T., and Cull-Candy, S.G. (1995). Intracellular spermine confers rectification on rat calcium-permeable AMPA and kainate receptors. J. Physiol. 486, 297–303. Keller, S., and Taylor, P. (1999). Determinants responsible for assembly of the nicotinic acetylcholine receptor. J. Gen. Physiol. 113, 171–176. Koh, D.S., Burnashev, N., and Jonas, P. (1995). Block of native

Assembly of AMPA and Kainate Receptors 113

Ca(2⫹)-permeable AMPA receptors in rat brain by intracellular polyamines generates double rectification. J. Physiol. 15, 305–312.

receptors is specified by two domains structurally related to bacterial amino acid-binding proteins. Neuron 13, 1345–1357.

Kunishima, M., Shimada, Y., Tsuji, Y., Sato, T., Yamamoto, M., Kumasaka, T., Nakanishi, S., Jingami, H., and Morikawa, K. (2000). Structural basis of glutamate recoginition by dimeric metabotropic glutamate receptor. Nature 407, 971–977.

Stern-Bach, Y., Russo, S., Neuman, M., and Rosenmund, C. (1998). A point mutation in the glutamate binding site blocks desensitization of AMPA receptors. Neuron 21, 907–918.

Kuusinen, A., Arvola, M., and Keinanen, K. (1995). Molecular dissection of the agonist binding site of an AMPA receptor. EMBO J. 14, 6327–6332. Kuusinen, A., Abele, R., Madden, D.R., and Keinanen, K. (1999). Oligomerization and ligand-binding properties of the ectodomain of the alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor subunit GluRD. J. Biol. Chem. 274, 28937–28943. Laube, B., Kuhse, J., and Betz, H. (1998). Evidence for a tetrameric structure of recombinant NMDA receptors. J. Neurosci. 18, 2954– 2961. Leuschner, W.D., and Hoch, W. (1999). Subtype-specific assembly of alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor subunits is mediated by their N-terminal domains. J. Biol. Chem. 274, 16907–16916. Liu, D.T., Tibbs, G.R., Paoletti, P., and Siegelbaum, S.A. (1998). Constraining ligand-binding site stochiometry suggests that a cyclic nucleotide-gated channel is composed of two functional dimers. Neuron 21, 235–248. Mano, I., and Teichberg, V.I. (1998). A tetrameric subunit stoichiometry for a glutamate receptor-channel complex. NeuroReport 9, 327–331. Mosbacher, J., Schoepfer, R., Moyer, H., Burnashev, N., Seeburg, P., and Ruppersberg, J. (1994). A molecular determinant for submillisecond desensitization in glutamate receptors. Science 266, 1059– 1062. O’Hara, P., Sheppared, P.O., Thogersen, H., Venezia, D., Haldeman, B.A., McGrane, V., Houamed, K.H., Thomsen, C., Gilbert, T.L., and Mulvihill, E.R. (1993). The ligand-binding domain in metabotropic glutamate receptors is related to bacterial periplasmic binding domains. Neuron 11, 41–52. Papazian, D. (1999). Potassium channels: some assembly required. Neuron. 23, 7–10. Partin, K.M., Patneau, D.K., Winters, C.A., Mayer, M.L., and Buonanno, A. (1993). Selective modulation of desensitization at AMPA versus kainate receptors by cyclothiazide and concanavaline A. Neuron 11, 1069–1082. Partin, K.M., Patneau, D.K., and Mayer, M.L. (1994). Cyclothiazide differentially modulates desensitization of AMPA receptor splice variants. Mol. Pharmacol. 46, 129–138. Pass, Y. (1998). The macro- and microarchitectures of the ligandbinding domain of glutamate receptors. Trends Neurosci. 21, 117–125. Premkumar, L.S., and Auerbach, A. (1997). Stoichiometry of recombinant N-methyl-D-aspartate receptor channels inferred from singlechannel current patterns. J. Gen. Physiol. 110, 485–502. Puchalski, R.B., Louis, J.-C., Brose, N., Traynelis, S.F., Egebjerg, J., Kukekov, V., Wenthold, R.J., Rogers, S.W., Lin, F., Moran, T., et al. (1994). Selective RNA editing and subunit assembly of native glutamate receptors. Neuron 13, 1–20. Rosenmund, C., Stern-Bach, Y., and Stevens, C.F. (1998). The tetrameric structure of a glutamate receptor channel. Science 280, 1596– 1599. Seeburg, P.H. (1996). The role of RNA editing in controlling glutamate receptor channel properties. J Neurochem. 66, 1–5. Sheng, M., and Pak, D.T. (2000). Ligand-gated ion channel interactions with cytoskeletal and signaling proteins. Annu. Rev. Physiol. 62, 755–778. Sommer, B., Keina¨nen, K., Verdoorn, T.A., Wisden, W., Burnashev, N., Herb, A., Kohler, M., Takagi, T., Sakmann, B., and Seeburg, P.H. (1990). Flip and flop: a cell-specific functional switch in glutamateoperated channels of the CNS. Science 249, 1580–1585. Stern-Bach, Y., Bettler, B., Hartley, M., Sheppard, P.O., O’Hara, P.J., and Heinemann, S.F. (1994). Agonist selectivity of glutamate

Swanson, G.T., Kamboj, S.K., and Cull-Candy, S.G. (1997). Singlechannel properties of recombinant AMPA receptors depend on RNA editing, splice variation, and subunit composition. J. Neurosci. 17, 58–69. Tu, L., and Deutsch, C. (1999). Evidence for dimerization of dimers in K⫹ channel assembly. Biophys. J. 76, 2004–2017. Uchino, S., Sakimura, K., Nagahari, K., and Mishina, M. (1992). Mutations in a putative agonist binding region of the AMPA-selective glutamate receptor channel. FEBS Lett. 308, 253–257. Yamada, K.A., and Tang, C.M. (1993). Benzothiadiazides inhibit rapid glutamate receptor desensitization and enhance glutamatergic synaptic currents. J. Neurosci. 13, 3904–3915. Washburn, M.S., Numberger, M., Zang, S., and Dingledine, R. (1997). Differential dependence on GluR2 expression of three characteristic features of AMPA receptors. J. Neurosci. 17, 9393–9406. Watkins, J.C., and Evans, R.H. (1981). Excitatory amino acid transmitters. Annu. Rev. Pharmacol. Toxicol. 21, 165–204. Wells, G.B., Lin, L., Jeanclos, E.M., and Anand, R. (2001). Assembly and ligand binding properties of the water-soluble extracellular domains of the glutamate receptor GluR1 subunit. J. Biol. Chem. 276, 3031–3036. Wenthold, R.J., Yokotani, N., Doi, K., and Wada, K. (1992). Immunochemical characterization of the non-NMDA glutamate receptor using subunit-specific antibodies. Evidence for a hetero-oligomeric structure in rat brain. J. Biol. Chem. 267, 501–507. Wenthold, R.J., Petralia, R.S., Blahos, J., II, and Niedzielski, A.S. (1996). Evidence for multiple AMPA receptor complexes in Hippocampal CA1/CA2 neurons. J. Neurosci. 16, 1982–1989.