Partial Agonism and Antagonism of the Ionotropic Glutamate Receptor iGLuR5

More than 50 structures have been reported on the ligand-binding core of the ionotropic glutamate receptor iGluR2 that belongs to the 2-amino-3-(3-hydroxy-5-methyl-4-isoxazolyl)propionic acid-type of receptors. In contrast, the ligand-binding core of the kainic acid-type receptor iGluR5 has only been crystallized with three different ligands. Hence, additional structures of iGluR5 are needed to broaden the understanding of the ligand-binding properties of iGluR5, and the conformational changes leading to channel opening and closing. Here, we present two structures of the ligand-binding core of iGluR5; one as a complex with the partial agonist (2S,3S,4S)-3-carboxymethyl-4-[(1Z,3E,5R)-5-carboxy-1-methyl-hexa-1,3-dienyl]-pyrrolidine-2-carboxylic acid (domoic acid) and one as a complex with the antagonist (S)-2-amino-3-[5-tert-butyl-3-(phosphonomethoxy)-4-isoxazolyl]propionic acid ((S)-ATPO). In agreement with the partial agonist activity of domoic acid, the ligand-binding core of the iGluR5 complex is stabilized by domoic acid in a conformation that is 11° more open than the conformation observed in the full agonist (S)-glutamic acid complex. This is primarily caused by the 5-carboxy-1-methyl-hexa-1,3-dienyl moiety of domoic acid and residues Val685-Thr690 of iGluR5. An even larger domain opening of 28° is introduced upon binding of the antagonist (S)-ATPO. It appears that the span of domain opening is much larger in the ligand-binding core of iGluR5 (30°) compared with what has been observed in iGluR2 (19°). Similarly, much larger variation in the distances between transmembrane linker residues in the two protomers comprising the dimer is observed in iGluR5 as compared with iGluR2.

The ionotropic glutamate receptors (iGluRs) 3 are responsible for the major part of the fast excitatory synaptic transmission in the mammalian brain. The iGluRs are divided into N-methyl-D-aspartic acid, 2-amino-3-(3-hydroxy-5-methyl-4isoxazolyl)propionic acid (AMPA), and kainic acid receptors based on selective agonist binding properties and sequence similarity of the receptor subunits (1)(2)(3). The three classes of iGluRs each consist of a number of subunits: NR1, NR2A-D, and NR3A-B for N-methyl-D-aspartic acid receptors, iGluR1-4 for AMPA receptors, and iGluR5-7 and KA1-2 for kainic acid receptors. The iGluRs share a similar tetrameric structure where the subunits assemble as a set of two dimers (4 -8). Each subunit contains a bi-lobed ligand-binding core attached to the transmembrane regions forming the ion channel pore. After the first crystal structure of the genetically engineered soluble form of the iGluR2 ligand-binding core (iGluR2-S1S2J) was reported, representative structures have been published for members of all three iGluR classes (9 -13). The majority of the structures, however, are co-crystals of iGluR2-S1S2J and agonists, which in combination with biophysical and biochemical analysis have provided molecular details of the receptor activation mechanisms. Surprisingly, only a few crystal structures have been solved of antagonist complexes (9, 12, 14 -16).
A number of iGluR agonists and antagonists have been developed by use of the isoxazole nucleus as a scaffold (3,29). The compound 2-amino-3-[5-tert-butyl-3-(phosphonomethoxy)-4-isoxazolyl]propionic acid (ATPO; Fig. 1) was developed as a hybrid structure based on the iGluR5 agonist 2-amino-3-(5tert-butyl-3-hydroxy-4-isoxazolyl)propionic acid and longchain phosphonic acid containing N-methyl-D-aspartic acid receptor antagonists such as 2-amino-7-phosphonoheptanoic acid. ATPO was first described in its racemic form as a competitive AMPA receptor antagonist (30). Upon resolution of the ATPO enantiomers, a more detailed study was performed of the pharmacology on recombinant AMPA and kainic acid receptors expressed in Xenopus laevis oocytes (31). In these studies, (S)-ATPO was shown to be an antagonist at iGluR1-4 (K i values ranging from 2.0 to 6.7 M) and less potent at iGluR5 (K i ϭ 23 M), but no activity was observed at iGluR6 or KA2. The R-enantiomer was virtually inactive at all subtypes (31). Important information on the binding mode of ATPO was obtained by determination of a crystal structure of (S)-ATPO in complex with the iGluR2 ligand-binding core (14). This study provided detailed information about iGluR antagonism and in particular showed how (S)-ATPO stabilizes an open form of the ligand-binding core.
In the present study, domoic acid and (S)-ATPO were selected as obvious ligands to gain further insight into the mechanism of kainic acid receptor versus AMPA receptor agonism and antagonism. At the same time, information on the mechanism of selectivity within the kainic acid receptor class was revealed. Thus, we present x-ray crystallographic data of iGluR5-S1S2 in complex with domoic acid and (S)-ATPO. These two structures, combined with the previously reported structures of iGluR5-S1S2 in complex with (S)-glutamic acid (10,11,16) and in complex with the antagonists UBP302 and UBP310 (16) allow a comparison of iGluR5 ligand-binding core domain opening and distances between the transmembrane TM1-TM2 linker amino acids of each protomer comprising the crystallographic dimer.

EXPERIMENTAL PROCEDURES
Materials-Chemicals were purchased from Sigma, unless otherwise specified. Domoic acid was obtained from Tocris Cookson (United Kingdom). (S)-ATPO was synthesized, purified, and resolved as previously described (30,31). The rat iGluR5-S1S2 construct was expressed and purified as previously reported (10).
Radioligand Binding Assay-For saturation binding, 50 ng of iGluR5-S1S2 protein was incubated with 0.065-18 nM (2S,4R)-[ 3 H]4-methylglutamic acid (47.9 Ci/mmol, Tocris Cookson, UK) for 2 h on ice in assay buffer (50 mM Tris-HCl, 10% (v/v) glycerol, pH 7.1, at 4°C). Nonspecific binding was determined in the presence of 1 mM (S)-glutamic acid. Competition experiments were carried out using 35 ng of iGluR5-S1S2 with 0.5-2 nM radiolabel in the presence of 0.01-1,000 nM domoic acid or 0.10 -500 M (R,S)-ATPO. Samples were filtered onto Schleicher & Schuell 0.2-m ME24 mixed cellulose ester filters and washed twice with 2 ml of ice-cold assay buffer. Radioligand binding to full-length iGluR5(Q) expressed in Sf9 insect cells was carried out as previously detailed (32). Radioactivity was determined by scintillation counting. Data were analyzed  Crystallization-A solution of 2.3 mg/ml iGluR5-S1S2 (in 10 mM HEPES, pH 7.0, 20 mM NaCl, and 1 mM EDTA) containing 5 mM domoic acid was used for the co-crystallization experiments, giving a protein to ligand ratio of 1:50. Crystallization was carried out at 6°C using the hanging drop vapor diffusion method. The drops contained 1 l of protein-ligand solution and 1 l of reservoir solution; the reservoir solution contained 15% polyethylene glycol 4000, 0.3 M lithium sulfate, and 0.1 M cacodylate, pH 6.5.
For co-crystallization with (S)-ATPO, a 6 mg/ml iGluR5-S1S2 solution (in 10 mM HEPES, pH 7.0, 20 mM NaCl, and 1 mM EDTA) was diluted 1:1 with ligand solution (10 mM HEPES, pH 7.0, 20 mM NaCl, 1 mM EDTA, and 40 mM (S)-ATPO) and used for crystallization experiments. Crystals were obtained at 6°C and at several conditions that all included polyethylene glycol 4000 and 0.3 M lithium sulfate but with buffers at different pH values. The crystal used for data collection was grown from 20% polyethylene glycol 4000, 0.3 M lithium sulfate, and 0.1 M cacodylate, pH 6.5.
Data Collections and Processing-Crystals of iGluR5-S1S2: domoic acid were flash cooled in liquid nitrogen after soaking in a cryo-protectant containing 20% glycerol in reservoir solution. X-ray diffraction data were collected at the BL2 beamline at BESSY, Berlin, Germany, equipped with a MAR345 image plate detector. The wavelength was 0.954 Å and the crystal diffracted to 2.5 Å. Data processing was performed using the HKL pro-  grams Denzo, XdisplayF, and Scalepack (34), and the CCP4 suite of programs (35). A data set on a flash cooled crystal of iGluR5-S1S2:(S)-ATPO was collected at the X11 beamline at DESY, Hamburg, Germany, to 1.85-Å resolution. The wavelength was 0.812 Å. The data were indexed and scaled using the HKL2000 package (34). Statistics of both data sets are listed in Table 1.
Structure Determinations-The iGluR5-S1S2:domoic acid structure was solved using molecular replacement employing the program AMoRe (36) from CCP4 (35). The crystal structure of iGluR2-S1S2J in complex with kainic acid (molA, Protein Data Bank entry code 1FW0) (9) was used as a search model. One clear solution containing two molecules was obtained. Model building was performed using the program Coot (37). Initially, an NCS averaged map was generated using CNS (38). In the first cycles of refinement, strict NCS was applied. All subsequent rounds of refinement were performed with restrained NCS. Parameter and topology files for the refinement of domoic acid were generated by the ProDrg server (39). The fully refined structure comprises Arg 432 -Lys 544 , the Gly-Thr linker and Pro 667 -Trp 799 (molA) and Arg 432 -Lys 544 , the Gly-Thr linker and Pro 667 -Cys 804 (molB). For statistics on refinements, see Table 1.
The structure of iGluR5-S1S2 in complex with (S)-ATPO was solved by molecular replacement using the program Phaser (40) with the structure of iGluR2-S1S2J in complex with (S)-ATPO (molA, PDB entry code 1N0T) (14) as search model. The program ArpWarp (41) was used to trace 96% of the residues. Further model building was done using the program Coot (37) cycled with refinement using the CNS package (38). The fully refined structure comprises Thr 433 -Gln 492 , Trp 498 -Lys 544 , the Gly-Thr linker, Pro 667 -Ser 711 , and Ser 715 -Gly 803 . For statistics on refinements, see Table 1.
Structure Analysis and Figure Preparation-The contacts program within CCP4 (35) was used to analyze the structures (hydrogen bonds and van der Waals interactions). Domain openings were calculated using DynDom (42). The dimerization interfaces were analyzed using the Protein-Protein Interaction Server (43). Figs. 2, 3, 5, and 6 were prepared with PyMol (44) and Fig. 4 with MOE (45).
Protein Data Bank Accession Numbers-The atomic coordinates and structure factor amplitudes of the iGluR5-S1S2:domoic acid and iGluR5-S1S2:(S)-ATPO structures have been deposited in the RCSB Protein Data Bank under accession numbers 2PBW and 1VSO.

RESULTS AND DISCUSSION
The ligand-binding core of iGluR5 was crystallized in complex with the agonist domoic acid and the antagonist (S)-ATPO, and the structures were determined to 2.5and 1.85-Å resolution, respectively. The binding affinities of these two ligands (K i ) for the iGluR5-S1S2 construct were determined by competition studies with [ 3 H]-(2S,4R)4methylglutamic acid ( Table 2). The K i values of domoic acid at iGluR5-S1S2 (5.56 nM) and at the full-length iGluR5(Q) receptor (1.11 nM) were found to be similar. The binding affinity of ATPO at iGluR5-S1S2 (19.0 M) seems to be lower than at the full-length iGluR5(Q) receptor (2.21 M) in this assay, but is in   Fig. 1. B, close-up view of the iGluR5-S1S2⅐domoic acid complex (molB) including potential hydrogen bonds within 3.0 Å (dotted lines). Bonds of protein and ligand are shown in green and yellow, respectively. Water molecules are shown as red spheres. C, close-up view of the selectivity determining loops of iGluR2 (red), iGluR5 (green), and iGluR6 (blue). This loop corresponds to the region around Arg 686 and Asp 687 of domain 2 in iGluR5. The structures of iGluR6-S1S2:domoic acid (PDB code 1YAE) (46) and iGluR2-S1S2J:kainic acid (PDB code 1FW0) (9) were superimposed onto the iGluR5-S1S2: domoic acid structure (domain 1 residues). In iGluR5 and iGluR6, domoic acid is colored green and blue, respectively. Kainic acid in iGluR2 is colored red. The chemical structure of kainic acid is shown in Fig. 1. agreement with previous studies (23 M) (31). Hill values of unity for all compounds indicate one homogeneous binding site population for both full-length iGluR5(Q) and iGluR5-S1S2.
In whole cell patch clamp experiments with a piezo-driven application system, domoic acid was a partial agonist at iGluR5 receptors heterologously expressed in CHO-K1 cells. 100 M domoic acid gave rise to ϳ40% (n ϭ 6) of control 10 mM glutamic acid currents (data not shown).
The Structure of iGluR5-S1S2 in Complex with Domoic Acid-In the crystal structure of iGluR5-S1S2 in complex with domoic acid, two molecules are present in the asymmetric unit of the crystal (molA and molB; Fig. 2). The electron density of domoic acid in the ligand-binding site of iGluR5-S1S2 is well defined and allowed unambiguous positioning of the ligand and identification of ligand contacts; see Fig. 3A. Domoic acid interacts with residues of iGluR5-S1S2 via hydrogen bonds, ionic interactions, van der Waals interactions, and water-mediated contacts. The potential hydrogen bonds/ionic interactions are listed in Table 3 and illustrated in Figs. 3B and 4A. The ␣-amino acid part of domoic acid interacts with residues Pro 516 , Thr 518 , Arg 523 , Ser 689 , and Glu 738 . Ser 689 , together with Thr 690 , is also involved in interactions with the acetic acid part of domoic acid. Furthermore, the carboxylate group on the hexa-1,3-dienyl moiety of domoic acid forms a hydrogen bond to the backbone nitrogen atom of Tyr 489 . A number of water-mediated contacts are seen between iGluR5-S1S2 and domoic acid. The acetic acid carboxylate group of domoic acid makes indirect TABLE 3 Interactions of the ligand-binding core of iGluR5-S1S2 with the agonist domoic acid and the antagonist (S)-ATPO Comparison with iGluR2-S1S2J in complex with (S)-ATPO and iGluR6-S1S2 complexed with domoic acid. Potential hydrogen bonds/ionic interactions to ligands within 3.3 Å are tabulated. Potential van der Waals interactions are given as Footnote a. a The following residues have the potential to form van der Waals interactions with ligand (within 5.0 Å): iGluR5-S1S2: ( The values in the iGluR2-S1S2J:(S)-ATPO (PDB code 1N0T) (14) and the iGluR6-S1S2:domoic acid (PDB code 1YAE) (46) are in both cases for molA. iGluR2 numbering is according to rat iGluR2 without signal peptide (Swiss-Prot accession code P19491). iGluR6 numbering is according to rat iGluR6 (Swiss-Prot accession code P42260). c Domoic acid molA and molB refer to the two molecules in the asymmetric unit of the crystal. d For atom numbering of (S)-ATPO and domoic acid, see Fig. 1. contacts to Thr 690 , Val 685 , Met 691 , Leu 736 , and Ser 741 through three water molecules ( Table 3). The previously shown water-mediated network (via three water molecules) from Ser 741 to (S)-glutamic acid observed in the iGluR5-S1S2:(S)-glutamic acid complex (10) is only partly conserved in the present domoic acid complex, because the water molecule interacting with the nitrogen atom of glutamic acid is displaced by the pyrrolidine ring of domoic acid (not shown).
The structure of iGluR5-S1S2 in complex with domoic acid has been compared with the structures of domoic acid in complex with iGluR6-S1S2 (PDB entry code 1YAE) (46) and kainic acid in complex with iGluR2-S1S2J (PDB entry code 1FW0) (9). A major difference between the ligandbinding cores of iGluR5 and iGluR6 versus iGluR2 is the loop region around Arg 686 and Asp 687 of domain 2 (Fig. 3C). This loop region is involved in binding of the hexa-1,3-dienyl moiety of domoic acid. Our observation that the conformations of this loop are conserved between iGluR5 and iGluR6, but differ from that of iGluR2, strongly support the suggestion by Nanao et al. (46) that this particular region may account for the selectivity of domoic acid toward kainic acid receptors. Minor differences in binding of the ␣-amino acid part of the ligands arise from differences in amino acid composition of iGluR2, iGluR5, and iGluR6. An additional hydrogen-bonding acceptor and donor is found in iGluR5 (Thr 518 ) and iGluR2 (Thr 480 ) compared with iGluR6 (Ala 518 ), resulting in potential hydrogen bonds between the Thr 518 side chain oxygen atom (OG-1) and the nitrogen atom of domoic acid in iGluR5 (molA, 3.3 Å; and molB, 3.2 Å; Table 3); and similarly between kainic acid and Thr 480 in iGluR2. The water-mediated network from domoic acid to iGluR5-S1S2 was not seen in the iGluR6-S1S2 complex that may be due to the lower resolution (3.1 Å) of the structure.
The Structure of iGluR5-S1S2 Complexed with (S)-ATPO-The crystals of iGluR5-S1S2 in complex with the antagonist (S)-ATPO contain one molecule within the asymmetric unit of the crystal. The electron density of (S)-ATPO is very well defined; see Fig. 5A. The amino acid moiety of (S)-ATPO is anchored primarily by a bidentate salt bridge between the carboxylate group and Arg 523 as well as by hydrogen bonds between the ammonium group and residues Pro 516 , Thr 518 , and Glu 738 (Figs. 4B and 5B; Table 3); a motif that is common for binding of iGluR ligands. The phosphonate group of (S)-ATPO is anchored to domain 2 by Ser 689 only; however, water-mediated contacts to Thr 690 and contact to a glycerol molecule are also observed.
In iGluR5-S1S2, the network of water molecules that connects Ser 741 to both the amino group and the distal carboxylic acid of (S)-glutamic acid through hydrogen bonds (10) is partially conserved in the antagonist bound form. In the (S)-ATPO complex, Ser 741 is connected to the nitrogen atom of the isoxazole ring through two water molecules, whereas a third water molecule seen in the (S)-glutamic acid complex is displaced by the tert-butyl group of (S)-ATPO. Although these two water molecules are conserved, their positions are slightly shifted due to the presence of the isoxazole ring (not shown).
Comparison of the binding mode of (S)-ATPO in iGluR5-S1S2 and in iGluR2-S1S2J (14) reveals a different location of the phosphonate group; see Fig. 5C, which results in slightly fewer direct contacts (Table 3). This may be caused by the fact that the iGluR5-S1S2:(S)-ATPO structure is more open than the iGluR2-S1S2J:(S)-ATPO structure and that a contact to a glycerol molecule is gained in the iGluR5 complex. Furthermore, in iGluR5-S1S2 the conformation of (S)-ATPO is stabilized via a water-mediated contact from the ␣-carboxylate group to the phosphonate moiety. The 3-[5tert-butyl-3-(phosphonomethoxy)-4-isoxazolyl moiety of (S)-ATPO functions as a spacer keeping domains 1 and 2 FIGURE 4. Two-dimensional ligand-receptor interaction plots of iGluR5-S1S2 in complex with domoic acid (A) and (S)-ATPO (B). iGluR5-S1S2 polar residues are shown as purple circles, acidic residues as purple/red, basic residues as purple/blue, and hydrophobic residues in green. Contacts from ligand to receptor side chains as calculated by the program MOE are shown as green arrows and from ligand to receptor backbone as blue arrows. Water molecules in contact with ligands are shown as white circles and their contacts as yellow lines. The extent of ligand and receptor exposure is shown as blue spheres differing in size. separated in iGluR5, thereby making it an antagonist as also previously observed in iGluR2 (14).
In addition to the present antagonist complex, structures of two compounds chemically distinct from (S)-ATPO (UBP302 and UBP310) in complex with a ligand-binding core mutant of iGluR5 have recently been reported (16). The amino acid part of (S)-ATPO and the two UBP structures bind in a similar way to the main anchor Arg 523 , and the tert-butyl group of (S)-ATPO takes up approximately the same space as the 5-methyl group on the pyrimidine 2,4dione ring of the two UBP structures (Fig. 5D). Furthermore, the ether oxygen atom of (S)-ATPO is positioned as is the keto group in position 4 of UBP. The carboxylate group on the distal thiophene points into a different area of the ligandbinding site in UBP302 and UBP310 compared with the phosphonomethoxy moiety of (S)-ATPO (Fig. 5D). Whereas the phosphonomethoxy moiety of (S)-ATPO is primarily anchored to Ser 689 , the distal carboxylate group of UBP is tethered to residue Thr 690 . Another difference is found in the conformation of the glutamic acid (Glu 738 in iGluR5) that anchors the ␣-amino group of the ligands. In the UBP complexes, this glutamic acid is modeled with two side chain conformations; both of these conformations being too far away from the ligands for hydrogen-bond formation. In the (S)-ATPO complex, Glu 738 makes an ion pair with the ligand.
(S)-ATPO displays selectivity toward AMPA receptors and iGluR5 and show no binding at iGluR6 (47). The most prominent difference between ligand-binding site residues of iGluR6 and iGluR1-4/iGluR5 is the presence of an alanine residue at position 689 in iGluR6, which is a serine residue at the corresponding position in the other receptors. This serine residue provides vital hydrogen bonds to the phosphonate group of (S)-ATPO via its side chain hydroxyl group, and such hydrogen bonds are abolished in iGluR6. Therefore, differences in the residue at position 689 (Ser/Ala) probably account for the marked selectivity profile.
It has previously been shown that domoic acid activates slowly desensitizing currents in iGluR5 channels, whereas (S)-glutamic acid activates rapidly desensitizing currents (49). To investigate interactions that serve to stabilize the dimer of the domoic acid complex relative to the (S)-glutamic acid complex (16), we performed a detailed analysis of FIGURE 5. Binding mode of the antagonist (S)-ATPO in iGluR5-S1S2. A, F o Ϫ F c omit electron-density map of (S)-ATPO contoured at 3. Oxygen atoms are colored red, nitrogen atoms are blue, the phosphorus atom is orange, and carbon atoms are yellow. The chemical structure of (S)-ATPO is shown in Fig. 1. B, close-up view of the iGluR5-S1S2⅐(S)-ATPO complex including potential hydrogen bonds within 3.0 Å (dotted lines). Bonds of protein and ligand are shown in green and yellow, respectively. Water molecules are shown as red spheres. C, superposition (on domain 1 residues) of the structures of (S)-ATPO in complex with iGluR5-S1S2 and iGluR2-S1S2J (molA, PDB code 1N0T) (14). Bonds of protein and ligand are shown in green and yellow for iGluR5-S1S2 and in gray and black for iGluR2-S1S2J. Water molecules were omitted for clarity. D, superposition (on domain 1 residues) of the structures of iGluR5-S1S2 in complex with (S)-ATPO and UBP310 (molA, PDB code 2F34) (16). Bonds of protein and ligand are shown in green and yellow for iGluR5-S1S2:(S)-ATPO and in gray and black for iGluR5-S1S2: UBP310. The chemical structure of UBP310 is shown in Fig. 1. the two structures (Fig. 6). As can be seen in Fig. 6A, the same residues are overall important for dimer formation but notably, the Arg 775 NH-2-Asp 776 OD-1 salt bridge and the Tyr 521 N-Glu 787 OE-2 hydrogen bond are only present in the domoic acid complex (Fig. 6, B-D) due to conformational differences. Arg 775 corresponds to the position of the R/G mRNA editing site previously demonstrated to affect desensitization of AMPA receptors (50,51). Furthermore, mutation of Asp 776 to a glycine residue in iGluR6 resulted in receptors with greatly accelerated desensitization kinetics relative to wild type receptors (52). Therefore, it can be hypothesized that the Arg 775 -Asp 776 salt bridge in the domoic acid-bound iGluR5 complex is involved in stabilizing the dimer in the active form and thereby being partly responsible for the slower desensitization kinetics. The backbone hydrogen bond of Tyr 521 to the side chain of Glu 787 is not present in the (S)-glutamic acid complex due to a small relocation of the individual monomers in this com-plex. Furthermore, Tyr 521 adopts different side chain conformations in the two structures (Fig. 6D). In AMPA receptors, the corresponding residue to Tyr 521 is a leucine (Leu 483 in iGluR2), and mutation of this single amino acid into a tyrosine prevents receptor desensitization (53). Recently, it was shown that cysteine substitution of Tyr 521 and Leu 783 in iGluR5 and corresponding residues in iGluR6 and iGluR7 creates kainate receptors locked in their active conformations by intermolecular disulfide cross-links (54). Thus, the position and conformation of Tyr 521 as well as the Tyr 521 N-Glu 787 OE-2 hydrogen bond may be another factor contributing to the slower desensitization kinetics of domoic acid.
A dimer similar to the one seen in the iGluR5-S1S2:domoic acid structure is formed by crystallographic symmetry in the structure of iGluR5-S1S2 in complex with the antagonist (S)-ATPO, but with a slightly smaller interface accessible area of 1019 Å 2 . In this complex, the Arg 775 -Asp 776 salt bridge and Tyr 521 -Glu 786 hydrogen bond are still present, whereas the hydrogen bond between Ser 761 and Gln 786 has disappeared (Fig. 6A).
Correlation between Domain Opening and TM1-TM2 Linker Distances-In iGluR2, a correlation has been found between the agonist efficacy and the degree of ligand-binding core closure that the agonists induce compared with the unbound open form (9,48,55). In contrast, antagonists stabilize the open form (9,14,15). As an apo structure is yet not available for iGluR5, it is not possible to calculate a corresponding domain closure in iGluR5. Instead, we have determined a domain opening relative to that of the full agonist (S)-glutamic acid in complex with iGluR5-S1S2 (PDB entry code 2F36, molA) (16), see Fig. 7A. A domain opening of 11°occurs upon binding of domoic acid in iGluR5-S1S2 and the opening is primarily caused by steric clashes between the 5-carboxy-1-methyl-hexa-1,3-dienyl moiety of domoic acid and residues Val 685 -Thr 690 of iGluR5. A similar domain opening of ϳ13°is seen in the structure of iGluR6-S1S2 with domoic acid (PDB entry code 1YAE) (46) relative to the (S)-glutamic acid-bound form (PDB entry code 1S50) (11).
An even larger domain opening of 28°is introduced upon binding of the antagonist (S)-ATPO in iGluR5-S1S2 (Fig. 7). This domain opening is only slightly smaller (1-2°) than the opening observed in the UBP302 and UBP310 structures with iGluR5 (16), which suggests that the antagonist structures of iGluR5 are good representatives for an apo iGluR5 structure. It appears that the span of domain opening is much larger in iGluR5 (30°) compared with what has been observed in iGluR2 (19°) (Fig. 7). For example, (S)-ATPO induces a 9°smaller domain opening in iGluR2-S1S2J compared with that seen in iGluR5-S1S2.
In Figs. 2 and 7A, the domain opening in iGluR5-S1S2 is correlated to the distance between the TM1-TM2 linker residues of the two protomers comprising the dimer. A similar correlation is also observed in iGluR2-S1S2J (Fig. 7B). However, much larger variation in the distances between TM1-TM2 linker residues are seen in iGluR5 compared with iGluR2. This suggests that iGluR5 undergoes a more drastic conformational change than iGluR2 upon activation by an agonist. Also, several D1-D2 interdomain contacts in iGluR5-S1S2 are gained upon binding of the full agonist (S)glutamic acid as compared with the partial agonist domoic acid and especially to the antagonist (S)-ATPO (Table 4). In the structure of iGluR5-S1S2 in complex with (S)-glutamic acid, in total 12 interdomain hydrogen bonds/ionic interactions are seen of which three are conserved among all three structures (Glu 738 -Tyr 764 , Thr 740 -Tyr 764 , and Thr 740 -Trp 799 ). The previously described interdomain "lock" between Glu 441 -Ser 721 in iGluR5-S1S2 (10) and between Glu 402 -Thr 686 in iGluR2-S1S2J (9) is not present in the complexes with domoic acid and (S)-ATPO. Instead, an interdomain contact is formed from the neighboring Glu 442 -Tyr 744 . Even though the number of iGluR5 complex structures is still sparse, the results suggest that a correlation exist between agonist efficacy at iGluR5 and TM1-TM2 linker distances (and domain closure) in the iGluR5-S1S2 construct as previously also observed for iGluR2 (48, 55).

FIGURE 7. Histogram of domain openings (filled bars) and distances between equivalent TM1-TM2 linker residues of the dimer (open bars) in
iGluR5-S1S2 (A) and in iGluR2-S1S2J (B) induced by different compounds. Domain openings were calculated relative to that of the (S)-glutamic acid complex. Distances between two Ile 668 in the iGluR5 dimer (Ile633 in iGluR2) are shown. In A, GLU, the full agonist (S)-glutamic acid (PDB code 2F36) (16); DA, the partial agonist domoic acid; ATPO, the antagonist (S)-ATPO; UBP310, the antagonist UBP310 (PDB code 2F34) (16); and UBP302, the antagonist UBP302 (PDB code 2F35) (16). In B, GLU, (S)-glutamic acid (PDB code 1FTJ) (9); CPW, the partial agonist CPW-399 (PDB code 1SYH) (48); KA, the partial agonist kainic acid (PDB code 1FW0) (9); APO, the apo structure (PDB code 1FTO) (9); and ATPO, (S)-ATPO (PDB code 1N0T) (14). In all cases, molA was used. Conclusion-Here, we have presented x-ray structures of the ligand-binding core of iGluR5 in complex with the partial agonist domoic acid and the antagonist (S)-ATPO. These structures add valuable information on the ligand-binding properties of iGluR5, and insight into the structural changes occurring upon agonist and antagonist binding. Of note, the span of domain opening is much larger in the ligand-binding core of iGluR5 than in iGluR2 and much larger variations in the distances between transmembrane linker residues between the two protomers comprising the dimer are seen in iGluR5. When comparing the structures of iGluR5 in complex with domoic acid and (S)-glutamic acid, conformational differences and differences in hydrogen bonding patterns between residues located at the dimer interface were seen, which might account for the different desensitization kinetics of (S)-glutamic acid and domoic acid. In addition, the selectivity profiles at the AMPA and kainic acid receptors were addressed. A major difference between the ligandbinding cores of iGluR5 and iGluR6 versus iGluR2 is the conformation of the loop region around Arg 686 and Asp 687 of domain 2. The observation that the conformation of this loop is conserved between iGluR5 and iGluR6, but not in iGluR2, strongly suggests that this difference may account for the selectivity of domoic acid toward kainic acid receptors. (S)-ATPO displays selectivity toward AMPA receptors and the kainic acid receptor iGluR5 and no binding at iGluR6. The most likely explanation for this selectivity profile is the presence of an alanine at position 689 in iGluR6, which is a serine in the four AMPA receptors and in iGluR5. The structural information gained is of importance for understanding mechanisms of agonist and antagonist binding at ionotropic glutamate receptors and will be valuable in future structure-based drug design.