Mutational Analysis of Kainate Receptors

Discussion
We have shown that two cloned goldfish kainate receptor subunits, GFKARa and GFKARb (15), can be effectively expressed in transfected mammalian cells (HEK 293 cells). The high level of expression permits detailed biochemical and pharmacological characterization and mutational analysis of these receptor proteins. This study revealed several important features of goldfish kainate receptors: (1) Using the ligand binding characteristics of GFKARa and GFKARb, several important determinants of agonist binding were identified. [3H]Kainate saturation binding analysis showed that GFKARa and GFKARb have relatively similar affinities, although GFKARb exhibited significant positive cooperativity that was absent in GFKARa. The affinity of GFKARa and GFKARb for various glutamatergic ligands (L-glutamate, quisqualate, domoate, AMPA and CNQX) were measured using a [3H]kainate binding inhibition assay. While GFKARa and GFKARb have similar affinity for domoate, CNQX, quisqualate, and AMPA, homomeric receptors formed from these two subunits showed a marked difference (> 30-fold) in the affinity for L-glutamate. Two conserved N-terminal sequence motifs important for ligand (particularly L-glutamate) binding were identified. Mutation of three residues (Q12, A51 and Y52) in the N-terminal extracellular region of GFKARb indicates that two conserved sequence motifs are in close proximity and are involved in forming the L-glutamate binding site. (2) A solvent accessible disulfide bond formed between two conserved Cys residues in the second extracellular region was identified. [3H]Kainate binding to native and expressed GFKARa and GFKARb was increased upon treatment with the reducing agent dithiothreitol (DTT), indicating that a solvent-accessible disulfide bond(s) is present in these receptor proteins. Mutational analysis of the three conserved extracellular cysteines of GFKARb indicated that Cys305 and Cys358 in the second extracellular region (between TMIII and TMIV) form a solvent-accessible disulfide bond and this disulfide bond is located outside the ligand binding cleft. (3) A deletion mutant (aTMII) of GFKARa, which lacks the segment previously thought to be the second transmembrane domain (TMII), was efficiently expressed in transfected HEK 293 cells. This mutant was N-glycosylated in the same manner as the wild type GFKARa and retains high affinity (KD of 66 nM) for kainate, indicating that kainate receptor subunit lacking the proposed "TMII," when expressed in transfected mammalian cells, can attain the wild-type topology and fold to form the ligand binding site. These data support a topology model with three transmembrane segments and are consistent with the notion that ionotropic glutamate receptors are modular proteins, in which ligand binding involves the movement of two large extracellular lobes (25).

Topological analysis of AMPA and kainate receptors showed that the N-terminal region preceding TMI and the region between TMIII and TMIV are extracellular (15, 21, 23-26; Figure 8A,). These two large extracellular regions are homologous to bacterial periplasmic amino acid binding proteins (27, 28), a class of structurally and functionally well characterized soluble proteins of the bacterial transport system. Preliminary models of the extracellular ligand binding domain of non-NMDA AMPA and kainate receptors have been presented based on homology modeling (31). The results of the mutational analysis of GFKARa and GFKARb in this study are largely consistent with the predictions of the model and assist its refinement as described below.

Figure 4: (A) Schematic of the proposed transmembrane topology of kainate receptor subunit. The waving line (a hair-pin loop) represents the TMII segment which is proposed to insert into the center of the transmembrane oligomeric complex. In the aTMII mutation this segment is deleted. (B) An enlarged view of the three dimensional homology model of the LAOBP-like extracellular ligand binding domain (31), which is formed by both the region preceding TMI and the region between TMIII and TMIV. N-glycosylation sites and a disulfide bond are shown. B1, B2, and B3 indicate three loops proposed to be part of the binding site. B1 contains the VTTILE motif, and B2 contains the DGKYG motif. The figure was produced from the coordinates using MOLSCRIPT (52) and RASTER3D (53). (C) Proposed interactions between L-glutamate and residues in the binding site. The figure was produced using Quanta (Molecular Simulations, Inc.). [click on image for higher resolution]

Two sequence motifs (VTTILE and DGKYG) in the N-terminal region are well conserved in iGluRs (Figure 2A,) and are likely to be involved in forming the ligand binding site. In bacterial periplasmic binding proteins, residues in these two regions are important for ligand binding. Y14 of lysine/arginine/ornithine-binding protein (LAOBP) and histidine-binding protein (HBP) as well as F52 of LAOBP and L52 of HBP directly contact the ligand (41-44). Y14 corresponds to the Glu of VTTILE, and F52 corresponds to the Tyr of DGKYG (28, 31). In both AMPA (38, 39) and NMDA (30) receptors, mutations of the position corresponding to Glu (VTTILE) and Lys (DGKYG) in these two motifs affect ligand binding. However, the effects are more dramatic at the Glu site than at the Lys site. A mutation of the Lys in DGKYG to Glu decreases affinity for L-glutamate only 20-fold, and the effects of mutations at the Lys site have differential effects on channel activation by kainate and L-glutamate (38, 39). GFKARb has uncharged groups in both the Glu and Lys positions (VTTIKQ and DGAYG) and binds L-glutamate with an affinity 30-fold lower than GFKARa (VTTIKE and DGRYG). Adding a negative charge to VTTILE (VTTIKQ to VTTIKE; Q12E mutant), increases the affinity, but adding a positive charge to DGKYG (DGAYG to DGKYG; A51K mutant) decreases the affinity. The double mutant has an affinity for L-glutamate only slightly different from the Q12E mutant. A thermodynamic analysis demonstrates that, for L-glutamate, the effects of the two mutations are not additive with respect to the double mutant. Clearly, A51 of GFKARb and R51 of GFKARa do not interact directly with the ligand in the binding site, as the affinity is decreased rather than increased in the b(A51K) mutation. Our model-building based on homology with bacterial periplasmic binding proteins indicates that these two motifs are in two short loops in proximity to the ligand binding site and are, thus, close to each other (31; Figure 8B,). The thermodynamic nonadditivity may indicate an interaction between the two positions. In the case of GFKARa and most other AMPA and kainate receptors, this would be an electrostatic interaction between the Glu of VTTILE and the Lys (or Arg) of DGKYG. Possibly, mutations in these two positions affect the subtle positioning of the two loops which in turn affects ligand binding. The effects of b(Y52F) and b(Y52S) indicate that the hydroxyl group of Y32 was not crucial, but the aromatic ring was important for ligand binding. Removal of the hydroxyl group, as in the b(Y52F) mutant, marginally reduced the affinity for kainate but not L-glutamate. However, replacement of the aromatic ring with a shorter chain, as in the b(Y52S) mutation, decreased affinity for kainate more than 100-fold. As mentioned above, the corresponding position in LAOBP and HBP directly contacts the ligand. One interpretation is shown in Figure 8C,, based on our previous model (31). We would propose that Glu (or Gln in the case of GFKARb) of the VTTILE motif interacts directly with the amine of L-glutamate. The tyrosine of DGKYG motif would then interact directly with the methylene groups of the L-glutamate side chain. Finally, the lysine (or arginine) of the DGKYG motif could also interact electrostatically with E12. In this case, E12 would interact with both the positive charge of the ligand and the positive charge of the DGKYG motif. In GFKARb, where position 12 is a Gln, the A51K mutation would not be able to accommodate both the positive charge of the ligand and the positive charge introduced by the mutation.

Residues in the second extracellular region are also involved in ligand binding, as indicated by the chimera (29) between AMPA (GluR1) and kainate receptors (GluR6) and mutations of NMDA receptor (NR1, 30) and frog KBP (22). A "hot spot" in this region is located about 30 residues C-terminal of TMIII. In an NMDA receptor (NR1), V666A and S669G mutations significantly decreased the affinity for glycine. Frog KBP (S267G) mutation (corresponding to S670 of NMDAR1) decreased the affinity for kainate 3-fold. This is likely due to either the loss of one carbohydrate chain normally attached at Asn265 of the wild type frog KBP, the Ser267 to Gly mutation, or the combination of both structural changes. It is interesting to note that mutations (T120M and T121A) in the corresponding sites of the histidine-binding protein significantly decrease the affinity for histidine (44). Proteins with diverse biological functions have been shown to undergo a large hinged domain motion upon ligand binding (45). The two large extracellular regions of kainate receptors are not only homologous to bacterial periplasmic amino acid binding proteins but probably retain the salient structural features of these binding proteins; that is, ligand binding involves the closure of two independent lobes. This notion is consistent with the effects on ligand binding of the cysteine mutations (C307S and C358S) of GFKARb and the deletion mutation of GFKARa (aTMII). Two cysteine residues, located in the loop between two transmembrane segments (TMIII and TMIV), are conserved in all the members of iGluR superfamily. In the case of goldfish kainate receptors, DTT treatment resulted in a decrease of the dissociation rate of kainate. The b(C305S) and b(C358S) mutants have increased affinity and decreased dissociation rate for kainate; both are no longer sensitive to DTT, indicating that Cys305 and Cys358 form this DTT-sensitive disulfide bond. This is consistent with the finding that two cysteine residues (C744 and C798) at the homologous positions of the NMDA receptor (NR1) mediate redox effect on channel activity, where DTT treatment of NMDA receptors increases the current about 2-fold (46, 47). Since these two Cys residues are absolutely conserved among members of the iGluR family, this proposed disulfide bond may be a common structural feature present in all iGluRs. One half of this S-S bond is very close to TMIII, potentially constraining the movement of the second extracellular domain. Our modeling suggests that the disulfide bond in this position may restrict the closure of the two lobes upon agonist binding (31). Perhaps the cleavage of the disulfide leads allows the two lobes to close with fewer restrictions in the bound form, resulting in the decreased rate of dissociation of kainate. The successful expression of aTMII is important for two reasons. First, the "TMII" segment is dispensable with regard to the overall folding and expression of the receptor proteins, since aTMII protein was as efficiently expressed in transfected HEK 293 cells as wild type GFKARa, while retaining the ligand binding profiles of the wild type GFKARa. This demonstrates further that TMII segment is not a true transmembrane region and lends further evidence in support of a topology model with three transmembrane segments. Secondly, this finding is consistent with our previous proposal that TMII is in a loop structure inserted into the center of the oligomeric transmembrane complex due to subunit folding and assembly (15). The TMII segment is not a simple cytoplasmic element. Deletion of TMII in the aTMII mutant decreased the affinity for L-glutamate and other glutamatergic ligands without changing the overall ligand binding profiles relative to each other. A likely explanation is that deletion of TMII did not directly affect residues in contact with ligand but may affect the closure of the two extracellular lobes. Deletion of TMII region may conceivably affect the packing of transmembrane segments, which in turn may affect the relative movement of the two extracellular domains in the ligand binding process, resulting into the slight decrease of affinity for some ligands. This effect may also be understood in light of the characteristics of bacterial binding proteins, where although the hinge regions connecting two lobes do not directly contact ligand, mutations of the hinge region can impair the hinge motion as in the case of S92F mutant of HBP (44). The combination of model building based on homology with bacterial periplasmic binding proteins and detailed mutational analysis promises to yield a much clearer and detailed picture of the extracellular ligand binding domain, which is a key requirement for the understanding of ligand binding and channel opening of iGluRs.

The four cloned and expressed nonmammalian kainate receptors (frog KBP, chick KBP, and goldfish GFKARa and GFKARb) all show physiologically meaningful affinities for L-glutamate (KD's of 1 to 70 然). The concentration of L-glutamate in the synapse shifts between 1 然 under resting conditions and 1 mM at its peak (48). The physiological role of these nonmammalian vertebrate kainate receptors is still not clear, since ion channel activity has yet to be demonstrated in heterologous expression systems. The possibility exists, however, that suitable conditions for the reconstitution of the potential ion channel activity have not been found in these expression systems. Nevertheless, kainate receptors have been implicated in the process of synapse formation in the avian brain, as indicated by the finding that kainate receptor expression was significantly induced by an imprint stimulus in the duckling hyperstriatum ventral (13). In case of the "orphan" receptor GluRd, although heterologously expressed GluRd shows no ligand binding activity and no ion channel activity, GluRd has important physiological roles in the cerebellum parallel-fiber/purkinje cell synapses as shown by a transgenic experiment (6). Alternatively, to understand their physiological role in nonmammalian vertebrate brain, the overall synaptic architecture and characteristics may need to be taken into consideration. It is noteworthy that the binding of glutamate to the glutamate transporter, instead of the actual re-uptake of glutamate, affects the clearance kinetics of glutamate in the synaptic cleft (49-51). The above features of the glutamate synapse raise the possibility that, in addition to possible ion channel activity, nonmammalian vertebrate kainate receptors might also affect the clearance of the neurotransmitter L-glutamate.

The mutational analysis of kainate receptors from goldfish brain strongly supports the functional as well as sequence homology to bacterial periplasmic amino acid binding proteins. Homologies between these lower molecular weight glutamate receptors and the 100 kDa AMPA and kainate receptor subunits found in both vertebrates and invertebrates suggests that the structure of the ligand binding domain is likely to be highly conserved. Thus, the disulfide bonding pattern described for goldfish receptors is likely to be a common feature of all ionotropic glutamate receptors as are the residues important for ligand binding.

Title Page
Introduction
Experimental Procedures
Results: L-Glutamate Binding Site
Results: Disulfide Bond
Results: Deletion of TMII
Discussion
References
Acknowledgements