Domain organization and function in GluK2 subtype kainate receptors

Abstract

Glutamate receptor ion channels (iGluRs) are excitatory neurotransmitter receptors with a unique molecular architecture in which the extracellular domains assemble as a dimer of dimers. The structure of individual dimer assemblies has been established previously for both the isolated ligand-binding domain (LBD) and more recently for the larger amino terminal domain (ATD). How these dimers pack to form tetrameric assemblies in intact iGluRs has remained controversial. Using recently solved crystal structures for the GluK2 kainate receptor ATD as a guide, we performed cysteine mutant cross-linking experiments in full-length tetrameric GluK2 to establish how the ATD packs in a dimer of dimers assembly. A similar approach, using a full-length AMPA receptor GluA2 crystal structure as a guide, was used to design cysteine mutant cross-links for the GluK2 LBD dimer of dimers assembly. The formation of cross-linked tetramers in full-length GluK2 by combinations of ATD and LBD mutants which individually produce only cross-linked dimers suggests that subunits in the ATD and LBD layers swap dimer partners. Functional studies reveal that cross-linking either the ATD or the LBD inhibits activation of GluK2 and that, in the LBD, cross-links within and between dimers have different effects. These results establish that kainate and AMPA receptors have a conserved extracellular architecture and provide insight into the role of individual dimer assemblies in activation of ion channel gating.

Glutamate receptor ion channels (iGluRs) mediate excitatory transmission in the mammalian brain (1). Understanding how they transduce the binding energy of glutamate into rapid activation and desensitization of ion channel gating requires structural information. For nicotinic acetylcholine receptors and related members of the large family of cysteine (Cys) loop receptors (2–4), and for acid-sensing ion channel and P2X receptors which bind protons and ATP (5, 6), the extracellular LBDs exhibit 5- and 3-fold rotational molecular symmetry about a central axis centered on the ion channel pore. By contrast, iGluRs are tetrameric assemblies in which the extracellular amino terminal domain (ATD) and ligand-binding domain (LBD) both assemble as a dimer of dimers (7–10). Surprisingly, in the recently solved structure of a full-length AMPA receptor, the subunit pairs that form the ATD dimer assemblies swap partners in the LBD (11). Here we test whether the assembly principles observed in GluA2 apply also to kainate receptors, identify the interfaces which form the dimer of dimers assembly in both the ATD and LBD of GluK2, and examine the functional effects of cysteine mutant disulfide cross-links in the ATD and LBD dimer assemblies. Our results suggest that the LBDs in a homomeric glutamate receptor tetramer are not functionally equivalent and have implications for understanding how heteromeric receptors function in the brain.

Results

Crystal Structure of the GluK2 ATD Tetramer Assembly.

We previously crystallized the amino terminal domain of GluK2 in two different forms [Protein Data Bank (PDB) 3H6G and 3H6H] and solved structures of the ATD dimer assembly by x-ray diffraction (10). In both crystal forms the subunits in a dimer assembly show 2-fold molecular symmetry, with a buried surface area of 1540 Ų per subunit almost equally distributed between domains R1 and R2 (10). From each of these structures a tetrameric assembly composed of AB and A′B′ subunit dimers can be generated by application of crystal symmetry operations (Fig. 1A). In the resulting N-shaped structure, there is a global 2-fold axis of symmetry, with contacts between the pair of dimer assemblies mediated exclusively by residues at the lateral edge of domain R2. The GluK2 ATD N and C termini, defined by the CA positions of Thr2 and Glu384 of each subunit, form twisted parallelepipeds with dimensions of 102 × 68 Å and 51 × 63 Å, respectively (Fig. 1B). The buried surface area of the interdimer interface is 600 Ų per subunit and involves 2-fold symmetric contacts between helices H, I, and K, including a salt bridge between Lys188 and Glu219, a hydrogen bond between Lys188 and the main chain carbonyl of Ser241, and a potential cation π bond between Lys181 and Tyr240. van der Waals contacts are formed between the side chains of Lys188, Leu212, Met214, Thr-218, Glu219, Tyr220, and Tyr240 (Fig. 1C).

Fig. 1.

Crystal structure of the GluK2 ATD tetramer. (A) Molecular surface viewed from the top (Left) and after rotation by 90° (Right) with the four subunits colored individually. (B) Ribbon diagram showing how the GluK2 ATD AB and CD subunit dimers pack via the lateral edges of domain R2 to form a tetramer; the CA atoms of Thr2 and Glu384, which define the N and C termini of the ATD in each subunit, are indicated by colored spheres connected by dashed lines; black dumbbells indicate the CA positions of L151 in the AB and CD dimers. (C) Stereo view of the interaction surface connecting subunits B and D in the GluK2 ATD tetramer. Side chains that mediate intersubunit contacts are represented as sticks; orange spheres indicate the CA positions of Gly215. The view matches that in the right panel in A.

ATD Cysteine Mutants Map Dimer Interaction Surfaces in Full-Length GluK2.

To establish whether packing in full-length GluK2 matches that seen in the isolated ATD, we generated a series of cysteine mutants and tested these for oligomer formation by Western blot analysis under oxidizing conditions. The wild-type GluK2 construct contains 12 native cysteine residues (12), and in control experiments we observed weak cross-linking which probably would interfere with analysis of the effect of cysteines introduced to test mechanisms of GluK2 assembly (Fig. 2). Analysis of GluK2 ATD and LBD crystal structures reveals that the pairs Cys-65 and -316 in the ATD and Cys-719 and -773 in the LBD form disulfide bonds and that Cys-91, -199, and -432 are buried in the protein interior (10, 13). Mutation of all of the remaining five cysteine residues to their counterparts in the GluA2 AMPA receptor was required to eliminate spontaneous cross-linking (Fig. 2A); the mutations made were C540Y; C545V, an RNA editing site in M1 (12); C564S in the cytoplasmic linker connecting M1 and M2; and C827N and C840Q, two palmitoylation sites in the cytoplasmic C terminus (14). Functional assays on the resulting 5× cysteine (–) mutant expressed in Xenopus oocytes and HEK cells revealed that voltage-dependent polyamine block, attenuation of desensitization by the lectin Con A, and the kinetics and extent of desensitization were not different from wild type. For example, in rapid perfusion experiments, k_des = 129 ± 13 s⁻¹, and steady-state desensitization reached 98.4 ± 0.4% in 10 μM copper phenanthroline (CuPhen), with similar values of 132 ± 16 s⁻¹ and 97.2 ± 1.2% in 10 mM DTT (n = 6) (Fig. 2F). Both parameters are similar to previously reported values of 165 s⁻¹ and 99.1% for wild-type GluK2 (8, 15), although on average the peak current for the 5× cysteine (–) was about 5- to 10-fold smaller than for wild type.

Fig. 2.

GluK2 ATD cross-links created by cysteine mutagenesis. (A) The blots show (left to right) spontaneous oligomerization for wild-type GluK2; the C540Y/C545V/C564S 3× (–) triple mutant; and the 4× (–) mutant which contained in addition the exchange C827N. The additional change, C840Q, to produce the 5× (–) mutant yielded a construct which ran as a monomer with the same mobility as wild-type GluK2 under reducing conditions. (B) The L151C mutation in the AB subunit ATD dimer interface produced spontaneous dimer formation (control, Con), which was not enhanced by incubation with CuPhen (oxidizing condition, Ox) and which was reversed by incubation with 10 mM BME (reducing condition, Red). (C) The G215C mutant in the BD subunit interdimer interface also produced spontaneous dimer formation which was not enhanced by incubation with CuPhen. (D) The M214C/Y240C double mutant, designed to cross-link helices I and K in the BD subunit tetramer interface, also produced dimer formation. (E) The L151C/ G215C double mutant showed bands for both dimers and tetramers. (F) Outside-out patch responses for GluK2 5× (–) had kinetics similar to wild-type GluK2; in oxidizing conditions the current in CuPhen (red curve; monoexponential fit, k_des = 131 s⁻¹) was 88% of the peak current in DTT (blue curve; k_des = 117 s⁻¹), but on average, peak currents in CuPhen were 101 ± 14% of the amplitude of those in DTT. (G) In reducing conditions; 10 mM glutamate activated rapidly desensitizing currents for the L151C mutant (blue fitted curve; k_des = 100 s⁻¹). The peak amplitude was inhibited in oxidizing conditions (red fitted curve; k_des = 102 s⁻¹; current in 10 μM CuPhen was 49 ± 7% of that in 10 mM DTT, P = 3%, n = 6 patches). The inhibition was readily reversible (green fitted curve; k_des = 89 s⁻¹). (H) The G215C mutant also was reversibly inhibited in oxidizing conditions (red fitted curve; k_des = 63 s⁻¹); control and wash responses in DTT were similar (blue and green curves; k_des = 61 s⁻¹ and k_des = 60 s⁻¹, respectively). The inhibition was statistically significant; peak current in CuPhen was 55 ± 8% of the DTT current (P < 0.05, n = 6). (I) L151C-G215C double mutants were more strongly inhibited in oxidizing conditions (red fitted curve; k_des = 74 s⁻¹; peak amplitude in CuPhen was 35 ± 4% of that in DTT, n = 4). The inhibition was fully reversible (green fitted curve; k_des = 74 s⁻¹).

Using nomenclature developed for the full-length AMPA receptor, we will refer to the A′ and B′ peptide chains in the GluK2 ATD tetramer as subunits “D” and “C.” In the isolated GluK2 ATD crystal structure, AB or CD subunit dimer assemblies are stabilized by a hydrophobic surface in domain R2 in which the CA atoms of Leu151 in the two subunits in a dimer pair are separated by 6.8 Å (Fig. 1B). Consistent with ideal geometry for disulfide bond formation revealed by modeling cysteine mutant rotamers, the L151C mutant formed cross-linked dimers spontaneously, and the extent of dimerization was not increased by application of 10 μM CuPhen (Fig. 2B). In the GluK2 ATD tetramer, the interdimer interface is formed by helix I of subunit B, which projects into a pocket formed by helices I and K of subunit D in the adjacent dimer, and vice versa. Introduction of a cysteine at the base of helix I produced spontaneous GluK2 cross-linking for the G215C mutant, and the extent of dimerization was not increased by application of 10 μM CuPhen (Fig. 2C). Further evidence for a dimer of dimers assembly formed by contacts between the B and D subunits in full-length GluK2 was obtained using the double mutant M214C/Y240C (Fig. 2D), which links the top of helix H in subunit B with the base of helix K in subunit D (Fig. 1B). When the L151C and G215C mutations were combined to cross-link simultaneously the AB, CD, and BD subunit dimer interfaces, Western blots revealed the formation of tetramers as well as dimers (Fig. 2E). In all cases, samples reduced by incubation with 10 mM beta-mercaptoethanol (BME) migrated as monomers.

Functional Effects of ATD Cross-Links.

We used outside-out patch recording with rapid application of 10 mM glutamate to test for functional effects of ATD cysteine mutant cross-links and focused on the L151C and G215C mutants identified by Western blot analysis (Fig. 2). We applied glutamate every 10 s and alternated every 5–10 responses between reducing conditions with 10 mM DTT and test responses under oxidizing conditions with 10 μM CuPhen (16). Control experiments revealed that the GluK2 5× (–) construct had activation and desensitization kinetics similar to wild-type GluK2 and that responses to glutamate were not sensitive to redox conditions (Fig. 2F). The G215C mutant also showed rapid activation and nearly complete desensitization (Fig. 2G), but responses to glutamate showed reversible inhibition by CuPhen. After oxidation the peak current was 55 ± 8% (P < 0.05, n = 7 patches) of that in DTT (k_des 93 ± 10 s⁻¹ in DTT and 107 ± 14 s⁻¹ in CuPhen), slightly slower than the value of 165 s⁻¹ for wild-type GluK2 (16). The L151C mutant also produced redox-sensitive responses (Fig. 2H), such that glutamate-activated peak currents in CuPhen were reduced in amplitude to 49 ± 7% of those in DTT (P < 0.05, n = 5 patches). Again, desensitization was apparently unaffected (k_des 103 ± 15 s⁻¹ in DTT and 114 ± 21 s⁻¹ in CuPhen).

When the G215C and L151C mutants were combined to produce a covalently linked tetrameric ATD assembly, the degree of inhibition in CuPhen (peak current, 35 ± 4% of that in 10 mM DTT) was greater than for the single ATD mutants (Fig. 2I). In fact, the extent of inhibition was similar to the product for the single mutants, suggesting that the accumulated cross-links independently limit the ability of glutamate to activate the channel. The rate of desensitization for the G215C/L151C double mutant was slowed compared with both wild-type GluK2 and the individual point mutants [k_des 73 ± 19 s⁻¹ in DTT and 86 ± 18 s⁻¹ in CuPhen (n = 5)] but remained insensitive to the redox condition.

GluK2 Ligand-Binding Tetramer Assembly.

Earlier structural work had established that the GluK1 and GluK2 LBDs assemble as homodimers (17, 18) that can be stabilized in their active conformation by protein engineering (19, 20) and by domain 1 cysteine mutants (8). Insight into how the LBD dimers pack to form a kainate receptor tetramer remained elusive, and low-resolution single-particle EM image analysis for AMPA receptors gave contradictory results (21–23). The crystal structure of a full-length AMPA receptor now has provided the necessary higher-resolution template to test assembly mechanisms in kainate receptors (11). We tested for formation of cross-links at the interdimer surface of the tetrameric LBD assembly by introducing cysteine mutants at two different locations in domain 2 of the GluK2 LBD. Based on subunit terminology derived from the full-length AMPA receptor structure, mutants proximal to the global 2-fold axis of LBD symmetry will link the A and C subunits, whereas distal mutants cross-link the A and B or C and D subunits in adjacent dimer assemblies (Fig. 3A). In addition we combined mutations in the LBD with those in the ATD to test whether the subunit cross-over observed in the full-length GluA2 crystal structure also occurs in kainate receptors.

Fig. 3.

GluK2 LBD cross-links created by cysteine mutagenesis. (A) Molecular surface of the GluA2 LBD tetramer (11) with each subunit colored using the same scheme as for the GluK2 ATD tetramer shown in Fig. 1. (B) Ribbon diagram for a homology model of the GluK2 LBD tetramer showing cysteine mutants designed to cross the A and C subunits and the A and B subunits in a tetramer assembly. (C) Western blot analysis for cross-links connecting the GluK2 LBD tetramer interface. The K667C mutation, which cross-links the loop connecting helices F and G in the AC subunit interdimer interface, produced spontaneous dimer formation, which was slightly enhanced by incubation with CuPhen (Ox) and reversed by 10 mM BME (Red). The S669C mutation at the N terminus of helix G also produced cross-linking of subunits in the AC subunit interdimer interface. The A676C/G770C double mutant, which cross-links subunits A and B in the LBD tetramer, produced weak spontaneous dimer formation, which was strongly enhanced by incubation with CuPhen. A similar effect was observed for the S680C/N771C double mutant, which also cross-links subunits A and B. (D) The double mutant L151C/K667C, which cross-links both the AB and CD subunit ATD dimer interfaces and the AC subunit dimer interface in the LBD, produced spontaneous dimer and tetramer formation, which was reversed by 10 mM BME; the cartoon shows the intersubunit connections in this mutant.

The first two mutations tested, K667C and S669C, cross-link the N terminus of helix G in subunit A with its proximal dimer partner in subunit C (Fig. 3B) and are equivalent to K663 and I664 in the AMPA receptor GluA2 subunit (11, 16). The second set of mutants tested is distal to the global axis of symmetry and cross-links the C terminus of helix G in subunit A with the C terminus of helix K in subunit B (and potentially with their symmetry equivalents in subunits C and D) (Fig. 3B). The K667C and S669C mutants both formed disulfide cross-linked dimers spontaneously, and the extent of dimerization was not increased by application of 10 μM CuPhen (Fig. 3C). Cross-links also could be formed by the pair A676C and G770C and by the pair S680C and N771C which link helices G and K with their distal dimer partners, but, in contrast to results for proximal mutants, the formation of cross-links was very strongly stimulated by 10 μM CuPhen (Fig. 3C).

We next investigated if the symmetry switch between the ATD and LBD “layers” seen in the AMPA receptor GluA2 crystal structure is also a feature of kainate receptor architecture. In GluA2, the dimer pairs in the ATD layer are formed by subunits AB and CD, whereas the LBD dimers are formed by subunits AD and BC (11). Thus we combined single-interdimer mutations in the ATD and LBD layers of GluK2 in an attempt to cross-link tetrameric assemblies (SI Materials and Methods). The GluK2 L151C mutant, which cross-links the A and B or C and D subunits in the ATD, combined with the K667C mutant, which cross-links the A and C subunits in the LBD, produced both dimer and tetramer formation (Fig. 3D), consistent with a cross-over of the subunits forming dimer assemblies in the ATD and LBD extracellular layers of a GluK2 receptor assembly (Fig. 3D).

Subunit Nonequivalence in the Ligand-Binding Domain Tetramer.

The LBD cysteine mutants used for our experiments were designed to produce intermolecular cross-links based on the crystal structure of an AMPA receptor trapped in the resting state by a competitive antagonist (11). These mutants would be expected to interfere with movements produced by agonist binding and thus limit subsequent receptor activation or desensitization. In contrast to the similar effects observed for intradimer and interdimer ATD cross-links, we observed distinct patterns of inhibition produced by cross-links between different pairs of subunits in the GluK2 LBD tetrameric assembly, revealing a functional distinction between subunits located at proximal and distal positions relative to the overall 2-fold axis of the receptor.

In oxidizing conditions, disulfide bonds that formed between the K667C or S669C mutants in diagonally opposed A and C subunits, proximal to the global axis of symmetry, led to a partial but readily reversible inhibition of the glutamate-activated current (Fig. 4 A and B). On average, compared with responses recorded in DTT, glutamate-activated peak currents were reduced in amplitude to 35 ± 8% (P < 0.05, n = 6) for K667C and 56 ± 11% (P < 0.02, n = 7) for S669C. These diagonally cross-linked mutants also showed reduced desensitization, so that, when oxidized, the steady-state current was larger than control (Fig. 5). This effect was most profound in the S669C mutant for which the steady-state current was 7 ± 2% of the peak in DTT, whereas in CuPhen the steady-state current increased about 4-fold, so that the ratio of steady-state to peak current was 26 ± 5% (P = 0.016, n = 7 patches). Linked to this effect were reproducible changes in the rate of desensitization: k_des 140 ± 16 s⁻¹ in DTT and 249 ± 17 s⁻¹ in CuPhen for K667C; k_des 66 ± 10 s⁻¹ in DTT and 96 ± 24 s⁻¹ in CuPhen for S669C (Fig. 5). Overall, the behavior of receptors harboring diagonal cross-links was similar to that observed for the I664C mutant in GluA2 (16) and resembles responses produced by some partial agonists at kainate and AMPA receptors.

Fig. 4.

Effects of LBD cross-links on GluK2 receptor function. (A) The K667C mutant showed partial inhibition by oxidation. In this patch, the peak current amplitude was reduced from 153 pA to 30 pA, and the rate of desensitization was similar in DTT and CuPhen (k_des = 188 s⁻¹ and 152 s⁻¹, blue and red curves, respectively); the extent of desensitization was reduced in CuPhen, so that relative to the peak steady-state currents were 5.3 ± 0.7-fold larger than in DTT. K667C responses frequently showed rundown, as illustrated by incomplete recovery of the peak amplitude upon return to DTT (green curve; k_des = 187 s⁻¹). (B) The S669C mutant exhibited fast desensitization in reducing conditions (blue curve; k_des 72 s⁻¹) that was slightly, but significantly, accelerated in oxidizing conditions (red curve; k_des 100 s⁻¹). As for the K667C mutant, the response was only partially (39% in this example) inhibited by oxidation. Desensitization was blocked substantially, with the steady-state current increasing to 19% of the peak in oxidizing conditions, compared with 2% in 10 mM DTT. (C) Glutamate-activated currents for the S680C-N771C mutant were much more strongly inhibited in CuPhen (peak amplitude 9% in this example); the inhibition was rapidly and readily reversible, as shown by blue- and green-colored exponential fits for control and recovery responses, respectively.

Fig. 5.

Bar plots illustrating kinetic effects of cross-links in the LBD. (A) Oxidizing conditions producing cross-links between AC and AB subunits give rise to distinct patterns of current inhibition. The 5× (–) construct was not inhibited by oxidation. Intermediate inhibition was observed for the K667C and S669C mutants, and much stronger reduction of the current was seen for the S680C-N771C and the A676C-G770C double mutants; number of patches indicated above bars. (B) Oxidation produced small increases in desensitization rate for all mutants, but the effect was statistically significant only for the K667C and S669C mutants (P = 0.03 for each, randomization test). Only a subset of responses in CuPhen for S680C-N771C (three of six patches) and A676C-G770C (four of six patches) were large enough to measure the desensitization rate accurately. (C) The steady-state current was strongly and significantly increased in amplitude for K667C and S669C but not for the AB subunit cross-linking mutants. *, P, < 0.05.

Interdimer cross-links distal to the global 2-fold axis of pseudosymmetry at the AB or CD interfaces caused a more profound inhibition of responses to glutamate but no increase in steady-state current (Fig. 4C). Compared with responses in DTT, glutamate-evoked currents were 6-fold smaller in CuPhen for A676C-G770C (peak amplitude 16 ± 4%, n = 6) and 10-fold smaller in CuPhen for S680C-N771C (peak amplitude 8 ± 2%, n = 6). The large inhibition of responses to glutamate under oxidizing conditions unfortunately limited our ability to measure changes in the rate of onset of desensitization accurately, but when accurate measurement was possible, differences between these rates of responses in CuPhen and DTT were not statistically significant (Table S1).

Discussion

Our results establish that the tetrameric assembly of GluK2 kainate receptors closely resembles that of AMPA receptors. However, high-resolution crystal structures (24), functional cross-linking experiments (8, 25), and site-directed mutagenesis (19, 26) reveal that, despite their common fold, the LBDs of glutamate receptors exhibit striking functional variation and that this variation often is underpinned by exquisitely subtle structural changes (15, 27). The isolated ATDs of GluK2 and GluA2 also have similar structures, but much less work has been performed on this domain (9, 10, 28). We find that the interdimer interface in the GluK2 ATD tetramer is more tightly packed than that in AMPA receptors, with a buried surface of 600 Ų, whereas for the corresponding GluA2 structure (PDB 3H5V) the buried surface is only 400 Ų and decreases to only 330 Ų in full-length GluA2 (PDB 3KG2). This difference also extends to the more extensive ATD dimer interface between the AB or CD subunits, which has a buried surface area of 1,657 and 1,540 Ų in the two GluK2 crystal forms (10), whereas for full-length GluA2 and the GluA2 isolated ATD the buried area is only 1,334 Ų (9, 11, 28).

Very little is known about the role, if any, of the ATD in either AMPA or kainate receptor gating. Allosteric signaling by the ATD is well established in NMDA receptors (29–31) but not in non-NMDA receptors, with the possible exception of kainate receptor modulation by polyamines and protons (32). Our results reveal that cross-links both within and between ATD dimer assemblies, produced by cysteine mutants that form redox-sensitive disulfide bonds, modestly inhibit activation by glutamate. The mechanism underlying this effect is uncertain, but it is possible that the formation of cross-links causes subtle changes in the orientation of the ATD, which in turn propagates to the LBD. Surprisingly, in view of models for glutamate receptor desensitization based on the full-length GluA2 crystal structure, which predicts large changes in the distance within and between ATD dimers (11), the ATD cross-links reported here do not alter the extent or rates of onset and recovery from desensitization (Fig. S1). Our results show that although crystallographic studies provide a rich framework for understanding glutamate receptor function, many basic details about the role of the ATD in gating mechanisms remain to be determined.

By contrast, a large body of work on the LBD of kainate and AMPA receptors provides a rational framework for explaining the diverse effects of cysteine mutants. For both GluK2 and GluA2, the domain 1 Y490C/L752C mutants stabilize the active conformation of the LBD dimer assembly and block desensitization (8), whereas the domain 2 G725C and S729C mutants, which trap GluA2 in the desensitized state, abolish responses to glutamate (16, 25). Here we report two different classes of domain 2 cysteine mutants that inhibit responses to glutamate. The GluA2 full-length crystal structure reveals that these mutants map to a previously unknown interdimer interface formed at the base of helix G in the antagonist bound resting state. The inhibitory effect of these mutants is consistent with the intuitive notion that contacts mediated by domain 2 interfaces in the resting state are likely to be broken upon receptor activation (11). Surprisingly, the K667C and S669C mutations that link the proximal A and C subunits in the LBD tetramer produce partial inhibition of the glutamate-activated peak current but also reduce desensitization, leading to an increase in steady-state current, whereas the A676C/G770C and S680C/N771C mutations that link the distal subunits produce only a strong inhibition of activation.

What leads to partial inhibition and reduced desensitization observed for the K667C and S669C mutants? A clue comes from studies on native AMPA and kainate receptors for which partial agonists such as the 5-substituted willardiines and kainic acid exhibit a negative correlation between efficacy and the extent of desensitization, mirroring responses for the GluK2 S669C mutant (33–35). Crystal structures show that these partial agonists produce less LBD closure than glutamate. Interdimer cross-links at S669C and K667C also would be expected to restrict domain closure for the LBD of each subunit in a dimer pair, suggesting a related mechanism of action. Alternatively, the smaller peak current and reduced desensitization seen for the S669C mutant could result from changes in gating kinetics and possibly agonist affinity. However, we excluded this possibility because the rates of both deactivation and recovery from desensitization are not greatly accelerated in the S669C mutant (Fig. S1).

Single-channel recording indicates that individual iGluR subunits can activate sequentially following the binding of glutamate (36), but at present it is not known if the four subunits in a tetramer produce equal activation of ion channel gating. Incomplete inhibition of glutamate responses by the K667C and S669C mutants might occur because only the subunits proximal to the pseudo–2-fold axis of symmetry are cross-linked, leaving the distal subunits free to close when glutamate binds. Restriction of the movement of distal subunits by the A676C/G770C and S680C/N771C mutants inhibited responses to glutamate more profoundly, consistent with either a dominant role in driving channel opening or formation of cross-links for both the AD and BC dimer pairs. In a symmetrical dimer of dimers assembly, cysteines in helix K of both distal subunits should form cross-links with their subunit proximal partners. However, the GluA2 full-length structure has local asymmetry in the LBD, and only one pair of subunits is within disulfide-binding distance (11). Our data do not provide information about the conservation of this asymmetry in kainate receptors or about whether in a membrane-embedded receptor it is possible for simultaneous cross-linking to occur at both sites, but the small glutamate response observed in oxidizing conditions (16 ± 4% of that in DTT for A676C/G770C) is consistent with the asymmetry observed in the crystal structure. Our results thus raise provocative questions about whether ion channel gating triggered by activation of the proximal and distal LBD dimer pairs is functionally equivalent. The stronger inhibition observed for the A676C/G770C and S680C/N771C distal subunit mutants suggests that these pairs could provide the majority of the gating force to open the channel, but further work is required to test this possibility.

In conclusion, it seems probable that there is a hierarchy of intersubunit interfaces in AMPA and kainate receptors. In the resting state D2 interfaces formed between proximal and distal subunits in the LBD are likely to be weak, and we propose that this weakness is essential for efficient receptor activation, because movement of the D2 domains away from the central axis of the iGluR tetramer drives ion channel opening. In contrast, the D1 LBD interface is strong enough to remain intact during receptor activation but breaks when the receptor desensitizes. Within the ATD, the local dimer interface, which underlies the formation of ATD homodimers with micromolar or higher affinity (9, 10, 28), is likely to be the most stable assembly in the extracellular segment, consistent with its proposed role in assembly (37, 38). By contrast, the small footprint of the ATD dimer–dimer interface suggests a much weaker interface. In summary, our results establish the presence of specific, small interfaces in the extracellular domains of kainate receptors and indicate that their arrangement is likely to be the same as that in AMPA receptors. Movement of these interfaces has significant roles in receptor activation and desensitization. Further work using a combination of approaches will be required to establish the extent of subunit-specific domain movements in iGluRs and how this extent varies during the processes of activation and desensitization.

Materials and Methods

Molecular Biology and Biochemistry.

For this study we used the pRK5 expression vector encoding rat GluK2 (previously called “GluR6”) with the three RNA editing sites encoding VCQ and with a C-terminal His tag. Mutants were generated using overlap PCR and confirmed by double-stranded DNA sequencing of the amplified region. The numbering refers to the mature polypeptide chain following cleavage of the signal peptide (10). HEK-293T cells grown in T25 flasks were transfected using polyethyleneimine as described previously (10). Following solubilization with DDM, SDS/PAGE with 4–12% gels, and transfer to PVDF, blots were probed using anti GluK2/3 clone NL9 antibody (Upstate) and developed using the Amersham ECL Western blotting detection kit (GE Healthcare) and x-ray film (SI Materials and Methods).

Electrophysiology.

Wild-type and mutant GluK2 were expressed by transient expression in HEK-293 cells for outside-out patch recording with fast solution exchange as previously described (16). Typical 10–90% solution exchange times were 300 μs, as measured at the open tip of the patch pipette. Data were fit in Axograph and plotted in Kaleidagraph (Synergy Software). Statistical significance was assessed using a two-tailed randomization test using the program RANTEST (kindly provided by Remigijus Lape and David Colquhoun, University College London). Additional details are given in SI Materials and Methods.