Transmembrane Topology of the Glutamate Receptor Subunit

Ionotropic glutamate receptors mediate most rapid excitatory synaptic transmission in the mammalian central nervous system. These receptors are divided into a-amino-3-hydroxy-S-methyl-4-isoxazole propionate (AMPA), kainate, and N-methyl-D-aspartate receptors based on pharmacological and electrophysiological characteristics. Ionotropic receptor subunits are inte-gral membrane proteins that have been proposed to have a large extracellular ligand-binding N-terminal domain, four hydrophobic transmembrane domains, and an extracellular C-terminal domain. In this study we have shown that both AMPA receptor subunits (GluR14) and kainate receptor subunits (GluR6/7) are glycosylated in adult rat brain; however, the kainate receptor subunits are glycosylated to a greater extent. Ex- amination of the sequences of AMPA and kainate receptors revealed that kainate receptors have several additional consensus sites for N-linked glycosylation; in-terestingly, one of these is located in the proposed major intracellular loop of the receptor subunits. To test the proposed transmembrane topology model for these re- ceptors, we have used site-specific mutagenesis of the GluRf3 subunit to remove the consensus glycosylation site located within the proposed intracellular loop. Mutagenesis of this site demonstrates that it is glycosylated in transiently transfected human embryonic kidney cells, which express functional kainate receptors. Since N-linked glycosylation has only been found to occur on extracellular domains of plasma membrane proteins, these results suggest that the proposed transmembrane topology model for the glutamate receptor subunits is incorrect. Combining these results with other recent data, we have proposed an alternative transmembrane topology model. acids based on mature protein (13)) mutation. The fidelity of the mu- tation was confirmed by DNA sequencing. Dansient Expression of GluR6 in HEK-293 Cells-The cDNA encod- ing the GluR6 subunit was transiently transfected into human embryonic kidney (HEK-293) cells (American Type Culture Collection CRL 1573) using calcium phosphate coprecipitation, with 20 pg of DNA/ 10-cm culture dish (14). Biochemical analysis was performed 48 h fol- lowing transfection. The cells were collected in 50 rn phosphate-buff-ered saline (pH 7.4) containing 1 rn phenylmethylsulfonyl fluoride, rn benzamidine, 40 pg/ml leupeptin, 40 pg/ml antipain, 40 units/ml Trasylol, and 10 rn EDTA. The were then isolated by centrifugation (65,000 x g for 10 min), resuspended in buffer, and collected by centrifugation a second time.

Ionotropic glutamate receptors mediate most rapid excitatory synaptic transmission in the mammalian central nervous system. These receptors are divided into a-amino-3-hydroxy-S-methyl-4-isoxazole propionate (AMPA), kainate, and N-methyl-D-aspartate receptors based on pharmacological and electrophysiological characteristics. Ionotropic receptor subunits are integral membrane proteins that have been proposed to have a large extracellular ligand-binding N-terminal domain, four hydrophobic transmembrane domains, and an extracellular C-terminal domain. In this study we have shown that both AMPA receptor subunits (GluR14) and kainate receptor subunits (GluR6/7) are glycosylated in adult rat brain; however, the kainate receptor subunits are glycosylated to a greater extent. Examination of the sequences of AMPA and kainate receptors revealed that kainate receptors have several additional consensus sites for N-linked glycosylation; interestingly, one of these is located in the proposed major intracellular loop of the receptor subunits. To test the proposed transmembrane topology model for these receptors, we have used site-specific mutagenesis of the GluRf3 subunit to remove the consensus glycosylation site located within the proposed intracellular loop. Mutagenesis of this site demonstrates that it is glycosylated in transiently transfected human embryonic kidney cells, which express functional kainate receptors. Since N-linked glycosylation has only been found to occur on extracellular domains of plasma membrane proteins, these results suggest that the proposed transmembrane topology model for the glutamate receptor subunits is incorrect. Combining these results with other recent data, we have proposed an alternative transmembrane topology model. Glutamate receptors are essential to a wide variety of normal and disease-related processes in th.e central nervous system.
(K. W. R. and R. L. H.), MSTP Training Grant GM-07309 (to C. B.), and * This work was supported by the Howard Hughes Medical Institute a CIDA grant from the National Institutes of Health (to L. A. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Each subunit is proposed to contain a large extracellular ligand-binding N-terminal domain, four hydrophobic transmembrane domains, and an extracellular C-terminal domain. This proposed topology is based, in part, on hydrophobicity analyses of the amino acid sequences of glutamate receptor subunits and has been strongly influenced by topology models for nicotinic acetylcholine receptor and y-aminobutyric acid type A receptor subunits. However, little independent evidence is available to support this model, and in fact, recent findings have called into question the proposed membrane topology of the NMDA and AMPA receptor subunits. These studies have shown that the C-terminal domain of NMDARl is phosphorylated by protein kinase C in transfected human embryonic kidney 293 (HEK-293) cells as well as in primary neuronal culture (9), suggesting that this region is intracellular. Furthermore, immunocytochemical studies indicate that the C terminus of the GluRl subunit is intracellular (10,11). Clarification of the actual transmembrane topology of the ionotropic glutamate receptor subunits is important in determining the structure of these receptors and in investigating their function and modulation.
In this study we have investigated the N-linked glycosylation of these receptors to test the current models of transmembrane topology of the receptor subunits. Since N-linked glycosylation only occurs on extracellular domains of plasma membrane proteins, we used site-specific mutagenesis to examine whether a consensus site for glycosylation found within the proposed major intracellular loop of the kainate receptor GluR6 is actually glycosylated and therefore extracellular. Our results indicate that this site, Asn-720, is glycosylated, suggesting that the proposed transmembrane topology model of glutamate receptor subunits is incorrect. We propose an alternative transmembrane topology model for these subunits.
Site-directed Mutagenesis-The cDNA encoding the GluR6 subunit was subcloned into a mammalian expression vector under control of a cytomegalovirus promoter and subjected to site-directed mutagenesis (Bio-Rad) as previously described (12). The antisense mutagenesis primer 5' CTGCGTGAG'M'GACAG'M'CCG 3' was used with singlestranded DNA template to generate the N720Q (numbering of amino The abbreviations used are: AMPA, a-amino-3-hydroxy-5-methyl-4isoxazole propionate; NMDA, N-methyl-o-aspartate; HEK, human embryonic kidney; ConA, concanavalin A , PAGE, polyacrylamide gel electrophoresis. acids based on mature protein (13)) mutation. The fidelity of the mutation was confirmed by DNA sequencing.
Dansient Expression of GluR6 in HEK-293 Cells-The cDNA encoding the GluR6 subunit was transiently transfected into human embryonic kidney (HEK-293) cells (American Type Culture Collection CRL 1573) using calcium phosphate coprecipitation, with 20 pg of DNA/ 10-cm culture dish (14). Biochemical analysis was performed 48 h following transfection. The cells were collected in 50 rn phosphate-buffered saline (pH 7.4) containing 1 rn phenylmethylsulfonyl fluoride, 10 rn benzamidine, 40 pg/ml leupeptin, 40 pg/ml antipain, 40 units/ml Trasylol, and 10 rn EDTA. The membranes were then isolated by centrifugation (65,000 x g for 10 min), resuspended in buffer, and collected by centrifugation a second time.
Receptor Deglycosylation-Membranes prepared from HEK-293 cells transiently expressing either wild-type GluR6 or mutant GluR6 were resuspended in the glycosidase reaction buffer containing protease inhibitors. The membranes were denatured by boiling in 1% SDS, and then octyl-ED-ghcopyranoside was added to a final concentration of 2.0% octyl-0-D-glucopyranoside, 0.1% SDS. The preparation was then incubated a t 37 "C with or without N-glycosidase F (6 unitdml) for 4 h. The same protocol was used for deglycosylation of the rat brain membranes, which were prepared as previously described (14).
Electrophysiology-HEK-293 cells were transiently transfected with cDNA encoding either wild-type GluR6 or GluR6 (N720Q), as described above. Forty-eight to seventy-two h following transfection, cells were transferred on glass coverslips to the stage of an inverted microscope for patch clamp recording in the whole-cell mode. Recordings were made at room temperature under voltage clamp with a holding potential of -60 mV. Currents were sampled a t 4 kHz, filtered a t 2 kHz, and were acquired and analyzed using PCLAMP and the Axopatch 200 amplifier (Axon Instruments). Rapid perfusion of agonist was accomplished using computer-triggered solenoid switches (Neptune Research) to control the flow of solution from the two sides of a theta tube (RED Scientific Glass Co.), as previously described (15).

Glycosylation of AMPA and Kainate Receptors in Rat
Brain-In order to characterize receptor glycosylation in vivo, crude synaptosomal membranes were prepared from adult rat brain and incubated with or without N-glycosidase F, a glycosidase that removes N-linked carbohydrates. The proteins were then resolved by SDS-PAGE and immunoblotted with antibodies specific for the AMPA and kainate receptor subunits (14-16). Although both AMPA receptor subunits GluR1-4 and kainate receptor subunits GluR6/7 have a predicted molecular mass of -99 kDa based on the amino acid sequence, AMPA receptor subunits in brain had an apparent molecular mass of approximately 106 kDa, while the kainate receptor subunits had an apparent molecular mass of 112 kDa. However, treatment of both AMPA and kainate subunits with N-glycosidase F reduced their apparent molecular mass to 99 kDa (Fig. 1). These results suggest that the GluR6/7 subunits have a higher level of glycosylation. It is important to note that the level of glycosylation of GluR1-4 and GluR6/7 in rat brain was the same as that observed with GluR1-4 (data not shown) and GluR6 subunits transfected into HEK-293 cells (see below).

Glycosylation of Recombinant GluR6 on
Asn-720Since the kainate receptor subunits are more highly glycosylated than the AMPA receptor subunits in vivo, we examined amino acid sequences to determine whether kainate receptor subunits had glycosylation consensus sites that were conserved in kainate but not AMPA receptor subunits. Several additional potential glycosylation sites are present in kainate receptors, and one of these, Asn-720, is especially intriguing since it is located in the proposed major intracellular loop of GluR6 ( Fig. 2A). To examine whether this site is actually glycosylated, we analyzed the glycosylation of wild-type and mutant GluR6 subunits transfected in HEK-293 cells. Glycosylation of wild-type GluR6 expressed in HEK-293 cells was similar to that observed in rat brain (Fig. 2B ). Interestingly, immunoblotting of these GluR6transfected cells with an antibody raised against the C terminus of GluR6 detected a 36-kDa breakdown product of the GluR6 protein that when deglycosylated with N-glycosidase F resulted in a 33-kDa protein fragment (Fig. 2B). Both the 36and 33-kDa proteins were present only in HEK-293 cells transfected with GluR6, and immunorecognition of this protein was blocked when transfected cell membranes were immunoblotted with R6/7 antibodies preadsorbed with excess peptide (Fig. 2 B ) . This result suggested that GluR6 contained a glycosylation site within the C-terminal36-kDa fragment. The only glycosylation consensus site present in GluR6 that was consistent with the apparent molecular mass and C-terminal location of the breakdown product was Asn-720.
To confirm that this Asn was the site of glycosylation we used site-directed mutagenesis to replace this Asn with a Gln (N720Q). Mutation of this Asn residue appeared to completely remove the N-linked glycosylation of the 36-kDa breakdown product, producing a protein with an identical molecular mass as the deglycosylated wild-type breakdown product (Fig. 2 B ) . N-Glycosidase F treatment of the mutant receptor did not affect the migration of the C-terminal breakdown product (Fig.  2B). In addition, a small change in the migration of the mature GluR6 (N720Q) protein, compared with wild-type GluR6, was consistently observed on 4% acrylamide gels (Fig. 2 0 . These results demonstrate that the elimination of the consensus glycosylation site at Asn-720 removes a small amount of the total glycosylation present on the mature GluR6 protein but completely removes the N-linked glycosylation present on the 36-kDa breakdown product.

HYAFMSSK PSALVKNNEEGIORTLTA D V A L L M E S T T I E V I T O R N C N L T O l G G L l
PNGase F --+ -+ test whether mutation of Asn-720 affected the overall structure, function, and surface expression of GluR6, we compared the electrophysiological properties of the expressed wild-type and mutant receptors (Fig. 3). Both wild-type and mutant GluR6 subunits produced functional kainate receptors that showed rapid desensitization to either 1 mM kainate (Fig. 3A) or 1 mM glutamate (Fig. 3, B and C). The apparent decay constants for receptor desensitization to rapid application of 1 mM glutamate were 5.5 2 1.2 ms (mean 2 S.D., n = 4) for wild-type GluR6 and 5. hippocampal neurons (18). These Cod-sensitive receptors have recently been identified as kainate-preferring glutamate receptors (19,20). Since ConA binds to and cross-links sugar moieties, its effect on receptor desensitization could be mediated through binding to glycosylated amino acid residues. We used the patch clamp technique to analyze the modulation of wild-type GluR6 and GluR6 (N720Q) ion channels by C o d treatment to determine whether glycosylation of Asn-720 is required for the modulation of desensitization by ConA. After incubating transfected cells with 5 1.1~ ConA for 2-5 min, desensitization was almost completely eliminated for both wildtype GluR6 (n = 30) and GluR6 (N720Q) (n = 4) (Fig. 3, B  and C). DISCUSSION Until recently it has been accepted that glutamate receptors, by analogy with other ligand-gated ion channels, are oligomeric complexes of homologous subunits and that each subunit contains four transmembrane domains, with both the N and C termini located extracellularly (Fig. 4A). A variety of data is consistent with certain aspects of this model. For example, there is substantial evidence that the proposed second transmembrane domain contributes to the function of the ion channel, since mutagenesis of residues in this region alters channel properties in AMPA, kainate, and NMDA receptor subunits (8). of the AMF' A receptor subunit GluRl alters the ligand binding properties of this subunit, confirming that this region is extracellular (21). Furthermore, studies indicate that Ser-684, located in the N-terminal half of the proposed major intracellular loop between TM3 and TM4 (Fig. 4), is phosphorylated by CAMP-dependent protein kinase in transfected mammalian cells (15,22), supporting the theory that this region is intracellular and accessible to protein kinases. However, other aspects of the proposed topology model have recently been disputed. Specifically, there is evidence that the C termini of NMDA and non-NMDA receptor subunits are intracellular (9-11). The C terminus of the NMDA NR1 subunit contains multiple phosphorylation sites that are phosphorylated both in vitro and in vivo by protein kinase C, demonstrating that this region is intracellular (9). In addition, immunocytochemical localization of GluRl using antibodies raised against the C terminus indicates that the C terminus is intracellular (10,ll).
The mapping of N-linked glycosylation sites is an effective way to determine which regions of plasma membrane proteins are extracellular (23). In this study, we have determined that Asn-720 of the kainate receptor subunit GluR6, which was previously thought to be intracellular, is glycosylated. Using site-directed mutagenesis, we replaced Am-720 with a Gln. This mutation reduced the N-linked glycosylation of the mature GluR6 subunit and completely abolished glycosylation of a 36-kDa C-terminal breakdown product of the GluR6 subunit. We have shown that the kainate receptor subunits are more highly glycosylated than the AMPA receptor subunits in brain and in transfected cells. It is interesting to note that ConA and

!tamate Receptor Subunit GluR6
other lectins eliminate desensitization of kainate receptors while they do not have large effects on desensitization of AMPA receptors, suggesting that the additional glycosylation of the kainate receptors may play a role in the modulation of desensitization of these receptors by ConA. However, our results show that mutation of one of these additional glycosylation sites on GluR6, Am-720, does not eliminate the modulation of GluR6 desensitization by ConA and suggest that a combination of glycosylation sites or sites other than Am-720 are important for this modulation.
These results demonstrate that Am-720 of GluR6, which originally was proposed to be part of the major intracellular loop, is in fact extracellular. As discussed above, previous studies have shown that Ser-684 of GluR6 is phosphorylated by CAMP-dependent protein kinase in transfected mammalian cells, confirming that this residue is intracellular (15,22). These data are consistent with a new proposed transmembrane topology (Fig. 4B) in which there is an additional transmembrane domain present in between the two domains referred to as TM3 and TM4, as has been recently proposed (8). This new proposed transmembrane domain, which we have called TM3a, must consist of a stretch of amino acids between 684 and 720. This model places the C-terminal half of the major intracellular loop outside of the cell and the C terminus inside the cell. If this structural model is valid for the AMPA receptor subunits as well, then alternatively spliced modules flip and flop in GluR14, which affect ligand binding and desensitization kinetics would be extracellular, as has been recently proposed (8).
The data in this study provide insight into the topology of the ionotropic glutamate receptor subunits, which is useful for firther investigation of the structure and functional regulation of glutamate receptors.