I. INTRODUCTION

Top Previous Next References

Fast excitatory neurotransmission in the mammalian nervous system is mainly mediated by glutamate. Albeit the relevance of this amino acid for brain function has been repeatedly underscored, it is impossible to overestimate it. On the one hand, glutamate acts as the transmitter at most excitatory synapses, and it is involved in the induction of long-lasting changes in the efficacy of neurotransmission, changes thought to be cellular correlates of memory formation. On the other, glutamate has a crucial role during the ontogeny of the nervous system, participating in the outgrowth of processes, in the formation and elimination of synapses, and in the activity-dependent fine tuning of exquisitely precise patterns of connectivity in several brain areas. Finally, alterations of glutamatergic neurotransmission have been related to the neuronal damage observed after episodes of ischemia and hypoglycemia, as well as to the etiology of a series of neurological conditions including epilepsy, Alzheimer's disease, Huntington's chorea, and amyotrophic lateral sclerosis. As a result of this multiplicity of functions, the number of studies devoted to the understanding of glutamate-mediated transmission has grown steadily over the years. Therefore, it is hardly surprising that ionotropic glutamate receptors, cationic channels responsible of converting the synaptic release of the amino acid into an immediate neuronal response, are among the most studied and best-understood molecules of the nervous system (77).

It is now well established that there exist three subtypes of ionotropic glutamate receptors, subtypes that have been named after their preferred ligands: -amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA)-preferring, N-methyl-D-aspartate (NMDA)-preferring, and kainate-preferring receptors. The cloning of a large number of glutamate receptor proteins and the discovery of their structural relationships has confirmed the legitimacy of this pharmacological subdivision. At the same time, it has paved the way to most of our current understanding of the biophysical properties and the physiological role of each subtype in the mammalian brain.

Nowhere is this fact better exemplified than in the study of kainate receptors (27, 51, 90-93). Although the early classification schemes of glutamate receptors, based solely of pharmacological and radioligand binding assays (112, 197), did postulate the existence of a separate class of molecules selective for kainate, the evidence supporting this idea was ambiguous. For example, the fact that a given neuron could exhibit a rapidly desensitizing response upon the application of AMPA (or quisqualate, another agonist commonly used in the early studies) and a nondesensitizing response to kainate was interpreted as an indication that each ligand was acting on a separate molecular entity. However, the cloning of GluR1, the first AMPA receptor subunit (68), led to the observation that homomeric channels made of this protein could be activated by AMPA and by kainate. Furthermore, the characteristics of the responses elicited by each ligand on GluR1 were very similar to those observed previously in nerve cells (15, 84, 87, 126). In this way, both agonists were shown to act on the same receptor protein, placing in serious doubt the mere existence of kainate receptors.

The subsequent cloning of additional glutamate receptor subunits led eventually to the unequivocal discovery of a group of true kainate receptors (10, 11, 40, 66, 185), channels with a strong preference for this agonist over AMPA, which displayed rapidly desensitizing responses, similar to those observed in preparations like the dorsal root ganglia (DRG) (71). Thus the cloning of a number of kainate receptor subunits was a real breakthrough in the study of these molecules and constitutes the foundation upon which the spectacular progress in their understanding has been built over the last decade.

In this review, we summarize our current knowledge on kainate receptors by first discussing their molecular, biophysical, and pharmacological properties, arenas where the insights derived from the analysis of AMPA and NMDA receptors have been a very valuable asset. We then highlight the possible roles of these receptor molecules during synaptic transmission and illustrate their relevance for brain function with some recent examples where the involvement of kainate receptors at the systems level has been demonstrated. Finally, we speculate on the possibility that this receptor subtype may participate in certain pathological conditions, mainly epilepsy, pointing to the necessity of developing more selective pharmacological agents to test the potential of kainate receptors as targets of antiepileptic drugs.

	II. MOLECULAR BIOLOGY OF KAINATE RECEPTORS

Top Previous Next References

Low-stringency, homology-based cloning studies performed during the early 1990s (10, 11, 40, 66, 114, 148, 185) led to the discovery of five kainate receptor subunits termed GluR5, GluR6, GluR7, KA1, and KA2, proteins with molecular masses of ~100 kDa. The degree of similarity among the primary structure of these proteins and other ionotropic glutamate receptor subtypes was shown to be low but significant: ~40% homologous to AMPA receptors and ~20% to NMDA receptors. Based on their amino acid sequence, kainate receptor subunits have been further subdivided into two classes. GluR5, GluR6, and GluR7 share a 75-80% homology, whereas KA1 and KA2 are 68% similar. In contrast, the similarity between the two subclasses is just 45%. A remarkable observation related to the existence of these two subgroups is that, in addition to their marked differences in primary structure, the same division can be traced based on two of their functional characteristics. First, the heterologous expression of GluR5-7 leads to the formation of homomeric channels that can be gated by kainate (40, 152, 160). KA1 and KA2, in contrast, do not form channels on their own but only when coexpressed with the other subunits (66, 185). The resulting receptors possess properties different from the homomers, revealing the ability of KA1 and KA2 to coassemble into functional heteromeric channels (Figs. 1 and 2). Second, binding assays on recombinant receptors have shown that KA1 and KA2 have a significantly higher affinity for kainate than GluR5-7. This observation has led to the suggestion that the two classes correspond to the high- and low-affinity kainate binding sites originally described by the early autoradiographic studies (112, 113). Although definitive proof is still missing, in situ hybridization studies have lent some support to this idea by showing that the distribution of the low- and high-affinity radioligand binding sites roughly predicts the observed localization of each subunit.

View larger version (30K): [in this window] [in a new window]   Fig. 1. A: membrane topology of the kainate receptor subunits. A single channel is believed to be composed of four subunits. Each subunit is ~900 amino acids long (~100 kDa) and spans the membrane three times (segments M1, M3, and M4) such that the NH2-terminal domain of the protein is extracellular and the COOH-terminal domain is intracellular. A fourth hydrophobic sequence (M2) forms the pore of the channel. This segment is thought to dip into the membrane forming a structure akin to a hairpin. The three highlighted residues correspond to sites susceptible to editing in the mRNA of GluR6 subunits. When the three positions are edited, the resulting receptors exhibit null permeability to calcium and a linear current-voltage relationship. B: homomeric GluR6 channels are sensitive to kainate but insensitive to DL--amino-3-hydroxy-5-methylisoxazole-propionic acid (AMPA). The currents in response to kainate (KA), glutamate (Glu), or domoate (Dom) desensitize very rapidly. AMPA fails to elicit any response in GluR6 homomers, but it acts on GluR6/KA2 heteromeric channels.

View larger version (18K): [in this window] [in a new window]   Fig. 2. Structure and function of GluR5 and GluR7 receptors. A: the rat GluR5 can be found as two different variants: GluR5-1, which contains 15 extra amino acids in the NH2-terminal region, and GluR5-2. GluR5-2 has been found to have three splice variants that differ in the sequence of their COOH-terminal domains. GluR5-2a is 49 amino acids shorter than the GluR5-2b because of the introduction of a premature stop codon. GluR5-2c contains an extra in-frame exon that makes it 29 amino acids longer than the GluR5-2b isoform. The COOH-terminal domain for GluR5-1 is as in GluR5-2b. B: typical responses induced by different agonists at homomeric GluR5-2a receptors. Kainate induces faster and deeper desensitization at heteromeric GluR5/KA2 receptors (unpublished data). [From Sommer et al. (160). Reprinted by permission of Oxford University Press.] C: GluR7 undergoes alternative splicing at the COOH-terminal domain. The insertion of a 13-amino acid segment out of the reading frame of the transcript leads to the generation of a protein 9 amino acids shorter (GluR7b). D: both isoforms can form homomeric channels sensitive to glutamate and kainate but insensitive to domoate and AMPA. Heteromeric GluR7/KA1, in contrast, responds to the application of AMPA. [Data from Schiffer et al. (152). Reprinted from Neuron with permission of Elsevier Science.]

In addition to the identification of these bona fide mammalian kainate receptors, several additional homologous proteins capable of binding kainate have been cloned from other species including chick (58), frog (73, 179, 184), and goldfish (192). These kainate binding proteins cannot form functional channels alone or when coexpressed with other subunits (65). It is likely that this phenotype is the result of their inability of translating ligand binding into channel opening. Evidence in favor of this idea has been obtained by the generation of functional chimeric receptors consisting of the GluR6 ligand binding domain and the ion channel domain of the different kainate binding proteins (177).

The analysis by in situ hybridization of the distribution of kainate receptors has revealed that they are widely expressed throughout the nervous system. However, the expression patterns of the different subunits are very heterogeneous (4, 191). Thus the GluR5 transcript is mainly present in DRG neurons, in the subiculum, in the septal nuclei, in the piriform and the cingulate cortices, and in Purkinje cells of the cerebellum (Fig. 3). GluR6 is most abundant in the cerebellar granule cells, in the dentate gyrus and the CA3 region of the hippocampus (Fig. 3), and in the striatum. The GluR7 mRNA is expressed at low levels throughout the brain but particularly in the deep layers of the cerebral cortex, in the striatum, and in the inhibitory neurons of the molecular layer of the cerebellum. KA1 is almost exclusively restricted to the CA3 region, although it is also expressed at lower levels in the dentate gyrus, in the amygdala, and in the entorhinal cortex. Finally, the KA2 message can be found in essentially every part of the nervous system. Although the different kainate receptor subunits are already present in the embryo, most of these patterns of expression emerge during the early postnatal period (4). Thus the amount of GluR5 mRNA peaks between postnatal day (P) 0 and P5 and then begins to decline toward its adult level by P12. Similarly, the GluR6 and KA1 expression patterns observed within the hippocampal formation begin to differ at around P0 (GluR6) and P12 (KA1).

It should be noted that the information gained by the in situ hybridization analysis, albeit correct in general, remains incomplete in the details. On the one hand, we still need to know what particular subunits are expressed by specific cell types within a given brain structure. For instance, a recent analysis of the coexpression of GluR5 or GluR6 with glutamic acid decarboxylase (GAD), the synthesising enzyme of GABA, revealed that many hippocampal interneurons contained both GAD and one of the kainate receptor subunits (122) (Fig. 3). The percentage of interneurons that expressed GluR5 and GAD plus those expressing GluR6 and GAD was larger than 100, suggesting that a significant fraction of individual cells expressed the two kainate receptor subunits. This finding led to the subsequent demonstration that members of the GluR5-7 subtype can coassemble to form heteromers (32, 122) (Fig. 4). On the other hand, the lack of subunit-specific kainate receptor antibodies has limited our understanding of their localization and compartmentalization within the nerve cell. The use of anti-GluR5/6/7 or anti-GluR6/7 antibodies has mapped kainate receptors to dendrites and postsynaptic membranes (56, 72, 132, 156). However, the advent of better antibodies against these proteins should help us to establish unambiguously where in the neuron each of them is localized and to determine what specific subunits mediate the different physiological effects of kainate receptor activation. In particular, the development of better antibodies against kainate receptors should help us to identify and dissociate between the effectors of its presynaptic versus its postsynaptic actions, as discussed in section IV.

View larger version (145K): [in this window] [in a new window]   Fig. 3. Expression pattern of GluR5 and GluR6 mRNA. In the cerebellum (A), GluR5 is expressed in the Purkinje cell layer (P.C.L.) and in Golgi cells of the granule cell layer (G.C.L.), whereas GluR6 is abundantly expressed by the granule cells (B). The asterisks indicate unlabeled Purkinje cell bodies. Insets show a magnification of the Purkinje cell layer. In the hippocampus, GluR5 is mainly expressed by interneurons (C), and GluR6 is predominantly expressed in the principal cells (D). Double in situ hybridization with glutamic acid decarboxylase (GAD) and either GluR5 (E) or GluR6 (F) probes is shown. The expression of GAD is labeled red (white arrowheads). The expression of GluR5 or GluR6 is labeled blue (black arrowheads). Coexpression appears as a brown precipitate (black arrows). E and F correspond to a close view of the CA1 and CA3 regions, respectively. S.O., stratum oriens; S.P., stratum pyramidale; S.R., stratum radiatum; S.L., stratum lucidum. [C-F from Paternain et al. (122).]

View larger version (43K): [in this window] [in a new window]   Fig. 4. Pharmacological and functional properties of some heteromeric kainate receptors. A: heteromeric GluR6/KA2 kainate receptors desensitize in response to glutamate but not to (RS)-s-amino-3(3-hydroxy-5-tert-butylisoxazol-4-yl)propanoic acid (ATPA), a drug previously thought to be a GluR5-specific agonist. In contrast, GluR5/KA2 heteromers fully desensitize in the presence of ATPA, whereas GluR5/GluR6 heteromers exhibit only a partial desensitization at high concentrations of this compound. B: dose-response curves of ATPA for GluR5/KA2 and for GluR6/KA2 heteromers. The EC50 values of the agonist for the two assemblies are 6.3 and 84 µM, respectively. C: dose-response curves of ATPA for GluR5/GluR6 heteromers. Preincubation with concanavalin A (ConA), a lectin that eliminates receptor desensitization, reduces the EC50 from 21 to 1.1 µM. [Modified from Paternain et al. (122).]

Kainate receptors are believed to share the same transmembrane topology and stoichiometry as AMPA and NMDA receptors (7, 67, 88, 143). Thus they are thought to be tetramers in which each monomer carries its own ligand binding site and contributes with a specific amino acid stretch to form the channel lumen, the so-called M2 segment, composed of hydrophobic residues that dip into the membrane forming a hairpin-like structure (Fig. 1A). In addition to this hydrophobic sequence, each subunit has three transmembrane segments (M1, M3, and M4) arranged such that the NH₂-terminal domain of each protein is extracellular and the COOH-terminal region lies intracellularly (139, 167, 192). By analogy to the structural data available for AMPA receptors (3), the glutamate binding site of kainate receptors is supposed to consist of the homologous residues distributed throughout both the loop between M3 and M4 and the NH₂-terminal domain (163). However, the actual structure of the binding pocket in kainate receptors is bound to be significantly different from the AMPA receptor. These likely differences should help to explain, among other things, why AMPA receptors can bind kainate with high affinity, whereas kainate receptors bind AMPA very poorly or not at all (Fig. 1B). In this regard, it is noteworthy that a specific residue of the extracellular loop between M3 and M4 has been identified as relevant for the moderate sensitivity of GluR5 to AMPA (165). When this amino acid is transplanted to the equivalent position in GluR6 (N721), the chimeric receptor can respond to the application of this agonist.

The identification of a number of isoforms derived from alternative splicing has revealed that the structural repertoire of the kainate receptor subtype is not limited to the five proteins mentioned earlier. The rat GluR5, for instance, can be found as two different variants (10): GluR5-1, an isoform that contains 15 extra amino acids in the NH₂-terminal region, and GluR5-2, a protein with three additional splice variants that differ in the sequence of their COOH-terminal domains. GluR5-2a is 49 amino acids shorter than the originally described GluR5-2 (now known as GluR5-2b) because of the introduction of a premature stop codon, whereas GluR5-2c contains an extra in-frame exon that makes it 29 amino acids longer than the GluR5-2b isoform (160) (Fig. 2A). Although it is possible that GluR5-1 also exists with three different COOH termini, the only isoform so far isolated presents the COOH terminus as in GluR5-2b. Similarly, two splice variants of the rat GluR7 have been discovered (Fig. 2C). In this case, the insertion of 40 out-of-frame nucleotides in the COOH-terminal sequence of GluR7b leads to the production of a slightly shorter protein that bears no significant homology in this region to any of the known ionotropic glutamate receptors (152). No alternative splicing has been reported for the rat GluR6, KA1, and KA2 subunits. However, the mouse GluR6 has been found to exist as two splice variants that differ in their COOH-terminal domains (59).

The role of the different kainate receptor splice variants is unknown. The 15 additional extracellular amino acids present in GluR5-1 do not map directly to the putative glutamate binding pocket of the receptor, but the possible influence of these residues on ligand binding has not been tested directly. Similarly, the possibility that this sequence may be involved in some type of intersubunit interaction remains unexplored. The rest of the known splice variants bear modifications in the COOH-terminal region. It is well established that the COOH-terminal domains of the different ionotropic glutamate receptor subtypes are important for their interaction with an increasing number of proteins involved in the organization and maintenance of synaptic structure. In the case of kainate receptors, this interaction is best exemplified by their ability to bind to members of the SAP90/PSD-95 class of proteins (54). Thus it is possible that the differences among the splice variants confer them with the selectivity necessary to interact with specific intracellular partners, providing the synapse with a system for the control of the activity of kainate receptors. This idea has not been explored in detail yet.

Another source of structural variability for kainate receptors is mRNA editing. As is the case for the GluR2 AMPA receptor (161), GluR5 and GluR6 (but not GluR7, KA1, and KA2) are susceptible of undergoing this kind of posttranscriptional modification at a specific position of their M2 segment, the so-called glutamine/arginine (Q/R) site. It has been shown that the presence of a R residue at this position in GluR2 is enough to change the rectification properties of AMPA receptors and, more importantly, to abolish their calcium permeability, a dominant trait observed in homomeric as well as in heteromeric AMPA channels (155, 161). In the case of homomeric kainate receptors assembled from GluR6(R), the Q-to-R substitution also decreases the permeability to calcium (20, 41) whilst increasing their chloride permeability (19). At the same time, the presence of a R at this site transforms the rectification properties of these receptors from inwardly rectifying to linear or slightly outwardly rectifying and reduces the unitary conductance of the channels (41, 86, 164). It should be noted, however, that the edition at the Q/R site is not the only change capable of altering the calcium permeability of GluR6 homomeric channels. Two additional positions prone to mRNA editing have been identified in the M1 segment of this subunit: the I/V site, where a valine can substitute for an isoleucine, and the Y/C site, where a tyrosine can be replaced by a cysteine (86). Editing at these positions modulates the effect of the Q/R site in calcium flow such that the fully edited subunit exhibits null passage of this cation. The mechanism of interaction among these three sites remains to be elucidated.

Heteromeric GluR5(Q)/GluR5(R) or GluR6(Q)/GluR6(R) receptors retain the rectification properties and the null permeability to calcium of the fully edited homomers (86, 160), indicating that the characteristics conferred by the presence of a R at M2 are dominant. In fact, this dominant character has recently been used to demonstrate that heteromeric receptors containing solely GluR5-7 subunits can exist. For instance, the coexpression of GluR6(Q) with GluR5(R) leads to the appearance of receptors with the characteristics of GluR6 channels but, at the same time, their current-voltage relationship is outward rectifying (32, 122). It must be noted, however, that not all of the functional changes introduced by mRNA editing are dominant as, for instance, GluR6(Q)/GluR6(R) channels remain chloride impermeable (19). Similarly, the reduction of the unitary conductance observed in the fully edited homomers is partly reverted in heteromeric assemblies. Thus the coexpression of KA2, a subunit unsusceptible to mRNA editing, with GluR5(R) or GluR6(R) generates receptors with unitary conductance larger than for the corresponding homomeric channels (70, 164).

As mentioned above, M2 is thought to form the pore of the channel. Therefore, it has been proposed that, in principle, the presence of a positively charged R within this segment is sufficient to explain the functional behavior of the edited receptor. First, the existence of a ring of positive charges at the pore reduces unitary conductance because of an increase in the energy barrier for the flow of positive charges, an increase with a more significant effect on divalent cations such as calcium. At the same time, the electrostatic cloud of positive charges could favor the passage of anions like chloride, although, in order for this to occur, it seems necessary to maximize the reduction of the energy barrier by editing all of the Q/R sites. Finally, the presence of an R residue at the pore prevents the entry of polyamines, molecules responsible for the inward rectification observed in kainate as well as in other receptors. This fact leads to the linear or to the outwardly rectifying current-voltage relationships observed in the edited channels (16, 19, 79). However, this model fails to account for the calcium permeability of GluR6 and its dependence on the editing at the M1 segment, as mentioned above.

Two additional facts about mRNA editing of kainate receptors deserve attention. First, the process is developmentally regulated, and it has been shown that a significant proportion of edited receptors exists both in embryonic and adult brains. The majority of GluR6 editing occurs earlier than for GluR5, and it is more thoroughly completed in the mature nervous system: up to 95% of GluR6 transcripts are edited, compared with 50-60% observed for the GluR5 subunit (9, 119-121, 153). Second, there seems to exist a site- and cell-specific regulation of kainate receptor mRNA editing. Studies of single hippocampal neurons in culture have shown that a given cell can coexpress several edited variants and, furthermore, that the degree of editing varies significantly across the different sites within the same neuron (144). Thus some cells have been found to possess, for instance, mRNA fully edited at the I/V site but totally unedited at the Q/R position. At the same time, only half of these transcripts appear edited at the Y/C site. Similar findings have been obtained in developing cerebellar granule cells (5) and in adult hippocampal neurons (102). The mechanisms responsible for the generation of this heterogeneity remain a matter of speculation and, similarly, its functional implications are not understood. For instance, mice with mutations at the Q/R editing site in GluR5 have been found to exhibit a reduction of kainate receptor-mediated currents in their sensory neurons (147), but the responses of these animals to painful stimuli are not altered. Furthermore, they do not show developmental abnormalities or deficits in the performance of several behavioral tasks. These observations stand in sharp contrast to the finding that knocking out the editing enzyme of the AMPA receptor subunit GluR2 leads to the generation of epileptic animals (45).

	III. FUNCTIONAL PROPERTIES OF KAINATE RECEPTORS

Top Previous Next References

The kinetic properties of the kainate receptor subtype have been studied mostly in heterologous expression systems. Under these conditions, stationary noise analysis has revealed that the single-channel conductance of homomeric GluR5(Q) or GluR6(Q) receptors is in the picosiemens range with values of 2.9 and 5.4, respectively (164). A detailed study of the resolvable channel openings has revealed the existence of three subconductance levels that are approximately twice or three times the size of the smallest current (164). It is interesting to consider these data in relationship to the tetrameric model proposed for AMPA receptors, where current through the channel is a function of the number of agonist molecules bound to the receptor (143). Do the subconductance levels observed in kainate receptors correspond to the opening of channels bearing a different number of ligand molecules, or is it the expression of a fundamentally distinct process? In any case, the existence of subconductance states may help to explain why studies performed under nonstationary conditions have obtained larger conductance values for GluR6 receptors. The larger conductance may reflect a preponderance of channels working at the higher subconductance levels (172).

As mentioned above, Q/R editing reduces the single-channel conductance of GluR5 and GluR6 receptors by more than one order of magnitude, an effect that is partly reversed by the coexpression of the KA2 subunit (70, 164). Thus the conductance of homomeric GluR5(R) channels is less than 200 fS, but it increases to 950 fS upon coassembly with KA2. Similarly, GluR6(R) homomers have a single-channel conductance between 230 and 260 fS, whereas the value for GluR6(R)/KA2 heteromers lies between 570 and 700 fS. The effect of KA2 on the conductance of homomeric GluR5(Q) or GluR6(Q) receptors is more subtle: an increase to 4.5 and 7.1 pS, respectively, which is accompanied, in the case of GluR5(Q), by a decrease in burst length (164). A paradoxical effect of the addition of KA2 subunits, the molecules with the highest affinity for kainate, to assemblies of GluR5 or GluR6 is that the heteromers exhibit a reduced affinity for the ligand (70, 164). This fact illustrates how little we understand about subunit interactions in kainate receptors and underscores other related questions that await an answer: how many KA1 or KA2 subunits can be incorporated into a receptor before it loses its functionality? Is it possible to find a subunit combination that has the highest affinity and the largest unitary conductance or does a compromise always exist between the two properties?

Some studies have tried to determine the single-channel conductance of kainate receptors expressed in neurons. In DRG cells, for instance, a value of 2-4 pS was determined (71) and three subconductance levels similar to those found for GluR5(Q) of GluR5(Q)/KA2 were observed. These findings agree with the fact that DRG neurons express GluR5 abundantly. In addition, the conductance of kainate receptors expressed in cerebellar granule neurons has been determined to be ~1 pS under conditions that reduce their desensitization (129) and ~4 pS in immature, proliferating cells (159).

Rapid desensitization is one of the most characteristic features of kainate receptors (Fig. 5A). The time course of current decay upon the constant presence of agonist follows a single exponential curve (94, 124), although double exponential decays have also been described (94, 144, 189). However, in addition to the fact that the speed of desensitization is dependent on the receptor subtype and on the cell type being analyzed, its actual value has been difficult to determine for different reasons (92). On the one hand, if the desensitization time constant is obtained purely from the measurement of current decay upon agonist perfusion, then a slow binding process could constitute a source of errors. On the other hand, the speed of current relaxation upon ligand application approaches the level of resolution of fast perfusion systems, raising the possibility that solution changes could also modify the measured desensitization rate. One way to circumvent these problems has been to measure the time constant of current relaxation at very high concentrations of agonist. Under these conditions, the binding rate is no longer a limiting factor and, as ligand concentration increases, the rate of current decay approaches an asymptotic value: the value of the real desensitization time constant. With the use of this strategy, it has been determined (124) that GluR6 homomers and native hippocampal kainate receptors have desensitization time constants of ~11-13 ms, a value similar to what has been found for recombinant channels in excised patches (63, 172) or lifted cells (164, 165). Whether the small difference that persists between the two measurements is due to an effect of the excision of the patch on the channel properties, as has been described for the NMDA receptor (151), remains to be determined.

View larger version (23K): [in this window] [in a new window]   Fig. 5. Desensitization properties of kainate receptors. A: the onset of kainate receptor desensitization is faster at higher concentrations of the ligand. At very high concentrations of the agonist, the constant rate of this process is approached by the asymptote of the rate of current relaxation. The desensitization rates for GluR6 homomeric channels and for native receptors of hippocampal neurons are virtually identical. The desensitization of GluR5/GluR6 heteromers occurs significantly faster. B: the exit from the desensitized state is much slower than its onset, and it is dependent on the nature of the agonist. Thus the recovery of the channels upon exposure to glutamate is completed in ~15 s. In contrast, ~1 min is necessary for complete recovery after exposure to kainate.

The recovery from the desensitized state proceeds slowly for kainate receptors, and it is dependent on the nature of the agonist (Fig. 5B). For instance, when glutamate is applied, two pulses evoke responses of similar sizes if they are separated by 15 s. In contrast, 1 min has to elapse when the desensitizing stimulus is a pulse of kainate (124). Another factor capable of affecting recovery from desensitization is the subunit composition (166). Thus GluR5 homomeric receptors recover in minutes, whereas the time constant of recovery for GluR5/KA2 heteromers is just 12 s. As mentioned earlier, the incorporation of KA2 subunits to kainate receptors reduces the affinity for the ligand. Thus it is possible that the faster recovery is due to a faster rate of agonist dissociation. However, it is unlikely that the exit from the desensitized state is a simple function of ligand unbinding because domoate, an agonist that desensitizes the receptors only partially, dissociates from the binding site with a faster time constant than the recovery time of the channel (94, 165). In any case, the profound differences in time scale between desensitization and recovery imply that the equilibrium between the two states is strongly displaced toward the former. In other words, the receptors spend most of their time in the desensitized state.

What are the implications of this propensity of kainate receptors to dwell on an inactive state? To visualize them more clearly, it is helpful to first look at the kinetic properties of their activation (Fig. 6). In cultured hippocampal neurons, glutamate has an EC₅₀ of 330 µM; 3 µM of this agonist does not induce a significant response, and 3 mM is a saturating concentration (124). In contrast, desensitization occurs at ligand concentrations two orders of magnitude lower, which elicit no activation of the channel (EC₅₀ = 2.8 µM). A similar situation has been observed in recombinant GluR6 homomers or when kainate is used as the agonist. In this case, the EC₅₀ for activation is 22 µM, whereas it is just 0.31 µM for inactivation (124). These observations imply that kainate receptors are very sensitive to resting levels of glutamate and, more importantly, that a large fraction of channels are already inactive before they can be activated, two factors that must have a profound influence in their physiological role. In addition, it is important to note that the kainate receptor activation and inactivation curves overlap to a significant extent (Fig. 6). This fact indicates that there exist concentrations of agonist (around 100 µM, in the case of glutamate) high enough to open the channels and, at the same time, low enough to produce only an incomplete desensitization. The current elicited by these concentrations of agonist has been termed "window current" by analogy to the one predicted by Hodgkin and Huxley during their study of the activation-inactivation curves of voltage-gated sodium channels (124). Although the window current remains to be directly measured, some physiological effects of kainate receptor activation lend indirect support to its existence, as will be discussed in section IVA (see Fig. 11C).

View larger version (24K): [in this window] [in a new window]   Fig. 6. Activation and inactivation properties of native kainate receptors. A: the current elicited by the application of a saturating concentration of kainate is substantially reduced by the administration of a previous pulse of glutamate in a dose-dependent manner. B: the open symbols correspond to the dose-response curve of receptor activation by glutamate. The solid gray symbols show the corresponding response to kainate after the receptors had been exposed to the indicated concentrations of glutamate. As can be seen, there is an interval of glutamate doses over which both curves cross over. This observation indicates that there are concentrations of the agonist high enough to open the channels and, at the same time, low enough to produce incomplete desensitization. The theoretical current predicted to exist under steady-state conditions of receptor activation is represented (dotted bell-shaped curve), and it reveals a steady activation of kainate receptors at 100 µM glutamate. [From Paternain et al. (124).]

A major hindrance in our way to understand kainate receptors has been the lack of specific agonists (Tables 1 and 2) and antagonists (Table 3). Although a clear pharmacological boundary has been traced between NMDA and the other ionotropic glutamate receptor classes, the separation between AMPA and kainate receptors has only been vaguely sketched and, for the longest time, the two types have been pooled together into what has been called the non-NMDA receptor subtype. For instance, AMPA has a negligible effect on recombinant kainate receptors (40, 66). It is inactive in GluR6 and GluR7 and has an EC₅₀ of 3 mM for the GluR5 subunit (160). Similarly, AMPA does not activate native kainate receptors in cultured hippocampal neurons or only at high concentrations in DRG cells (71, 94, 194). In contrast, kainate, albeit showing a clear preference for kainate receptors, has a very significant effect on AMPA channels at relatively low doses. The difference in EC₅₀ between the two receptors lies around a mere 5- to 30-fold higher affinity for kainate receptors (29, 71, 94, 188). At the same time, the classic non-NMDA receptor antagonists, 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), 6,7-dinitroquinoxaline-2,3-dione (DNQX), and 6-nitro-7-sulfamoylbenzoquinoxaline-2,3-dione (NBQX) (69), show at best a three- to fivefold selectivity between AMPA and kainate receptors (12, 98, 188). Thus the realization that kainate could activate both receptor subtypes, and that the classic battery of competitive antagonists was also unable to distinguish between them, led the field to a stagnation that lasted until the mid-1990s, a state that, fortunately, has begun to change.

The affinity of kainate for its different receptor subunits is very variable (Table 2). As mentioned earlier, radioligand binding assays have identified two subclasses of kainate receptors with different affinity. The high-affinity subunits KA1 and KA2 have a dissociation constant (K_d) of ~4-15 nM (66, 191). This figure is one order of magnitude higher than for GluR5-7, the low-affinity subunits (11, 160) (K_d = 50-100 nM). However, functional studies on recombinant receptors have revealed that the actual effective concentrations of kainate are much higher and vary depending on the subunit. Thus the EC₅₀ for the GluR5 subunit (160) is 33.6 µM, 299 µM for the GluR6 subunit (164) (although other values have been reported) (78), and >1 mM for the GluR7 subunit (152). A similar situation has been found for glutamate, the endogenous agonist (Table 2). These figures illustrate that kainate receptors do not have a high affinity for their prototypical activators. This conclusion can be extended to the native channels because the EC₅₀ values for kainate and glutamate measured in hippocampal and DRG neurons are similar to those obtained on some recombinant receptors (94, 124, 189).

In addition to kainate and glutamate, several molecules have been shown to activate kainate receptors with a certain degree of specificity. Domoate, one of the first to be identified, is 20- to 25-fold more effective than kainate on DRG cells and on recombinant GluR5 subunits (71, 160), although it is inactive on GluR7 homomers (152). It induces a slowly desensitizing response, and it shows a 50-fold preference for kainate over AMPA receptors (71). This selectivity has been now largely surpassed by a new generation of agonists among which (RS)-s-amino-3(3-hydroxy-5-tert-butylisoxazol-4-yl)propanoic acid (ATPA) is the prime example (89). The EC₅₀ of ATPA is 0.6 µM for native kainate receptors in DRG neurons and 2.1 µM for recombinant GluR5 subunits (29), an order of magnitude more effective than kainate. At the same time, its affinity for AMPA receptors is about 500 times lower (340 µM in cerebellar Purkinje cells). Nevertheless, these figures must be interpreted with caution because they have been measured from receptors treated with concanavalin A, a molecule that reduces the desensitization of the receptor and artificially increases its affinity (see below). In addition, it must also be noted that ATPA has been hitherto regarded as a GluR5-specific agonist. It is inactive on GluR6 homomeric channels, and it exhibits a potency 1,000-fold weaker for displacing [³H]kainate from GluR7 than from GluR5 in binding assays (29). However, it has been recently shown (122) that ATPA can also act on GluR5/KA2 (EC₅₀ = 6.3 µM), GluR6/KA2 (EC₅₀ = 84 µM), and GluR5/GluR6 heteromers (EC₅₀ = 21 µM). Furthermore, each of these assemblies exhibits a different desensitization profile in response to the agonist (Fig. 4). These findings may lead us to revisit previous findings where the selective activation of GluR5 is presumed to be responsible for the presynaptic effects of hippocampal kainate receptors, as discussed later.

Some derivatives of willardiine have also been advanced as selective kainate receptor agonists. Among them, (S)-5-trifluoromethylwillardiine is the most effective (EC₅₀ = 74 nM on DRG cells), but its selectivity for kainate over AMPA receptors is only 50-fold, a relatively low figure (194). On the other hand, (S)-5-iodowillardiine shows a ~130-fold selectivity for kainate receptors, but its affinity is lower (194) (EC₅₀ = 140 nM). As is the case for ATPA, (S)-5-iodowillardiine produces desensitizing currents on GluR5 but not on GluR6 or GluR7 homomers. In fact, it has been demonstrated that a single amino acid in the loop between M3 and M4 confers GluR5 with this differential sensitivity (166). However, this observation fails to explain why GluR6/KA2 heteromers are also responsive to this molecule (166). Finally, it is noteworthy that replacing iodine for fluoride transforms this drug into an AMPA receptor agonist with a 50-fold selectivity over kainate receptors (61, 194).

SYM 2081, the (2S-4R) diastereomer of 4-methylglutamate, is one of the latest additions to the list of specific kainate receptor agonists. It displays a selectivity three orders of magnitude larger for kainate than for AMPA receptors both in binding and in functional assays, but the selectivity of this molecule for kainate over NMDA receptors is significantly lower, only 200-fold (60). Although its pharmacological profile is incomplete, SYM 2081 does not seem to show the subunit specificity observed for ATPA and for (S)-5-iodowillardiine, as it elicits rapidly desensitizing currents on GluR5 and GluR6 homomeric channels (38, 78, 189, 198). Lastly, SYM 2081 has also been used as a functional antagonist of the kainate receptors (96). The activation and desensitization curves of GluR6 channels in response to this agonist (78) and to glutamate (124) show a minimal overlap. Therefore, its presence at low concentrations drives the receptor to an inactive state, preventing its subsequent opening by other agonists.

The quest for specific kainate receptor antagonists has not been as successful as the search for agonists. As mentioned above, the first generation of non-NMDA receptor blockers such as CNQX, DNQX, and NBQX cannot distinguish between kainate and AMPA receptors. The molecule 5-nitro-6,7,8,9-tetrahydrobenzo[g]indole-2,3-dione-3-oxime (NS-102) was then proposed as a selective kainate receptor antagonist based on its ability to block low-affinity [³H]kainate binding to cortical membranes (76). However, functional assays have yielded contradictory results regarding its specificity for kainate over AMPA receptors. Thus it was first shown that NS-102 blocked recombinant GluR6 and kainate receptors at a concentration 20 times lower than required for AMPA channels (173). However, it was later found that the IC₅₀ values of this antagonist for kainate and for AMPA receptors in cultured hippocampal neurons was very similar (2.2 and 4.1 µM, respectively). More importantly, ~30% of the current through either channel subtype remained unaffected at the solubility limit of NS-102, a fact that has greatly constrained its usage (125, 188).

The development of a new series of decahydroxyisoquinoline carboxylates appears to indicate the path to follow in the development of highly specific kainate receptor antagonists. The molecule LY293558 [(3S,4aR,6R,8aR)-6-(2-(tetrazol-5-yl)ethyl)decahydroxyisoquinoline-3-carboxylate], a noncompetitive antagonist of AMPA receptors (IC₅₀ = 0.5 µM), also blocks GluR5-mediated responses (IC₅₀ = 2.5 µM) with no effect on GluR6 homomers (13) (IC₅₀ >300 µM). A related derivative, LY294486 [(3SR,4aRS,6SR,8aRS)-6-((((tetrazol-5-yl)methyl)oxy)methyl)-1,2,3,4,4a,5,6,7,8,8a-decahydroxyisoquinoline-3-carboxylate], has a significantly higher selectivity for GluR5 than for AMPA receptors (IC₅₀ = 4 µM and 30-100 µM, respectively) and no effect on GluR6 or GluR7 homomeric channels (29). Lastly, the compound LY382884 [(3S,4aR,6S,8aR)-6-(4-carboxyphenyl)methyl-1,2,3,4,4a,5,6,7,8,8a-decahydroxyisoquinoline-3-carboxylate] has an even higher selectivity for GluR5 over the other kainate receptor subunits and, more importantly, over AMPA receptors (14, 157).

Despite this progress, the real breakthrough in the pharmacological distinction between AMPA and kainate receptors has been the advent of a series of 2,3-benzodiazepines that prevent the activation of AMPA receptors in a selective and noncompetitive manner (178). Among them, GYKI 53655 (also known as LY300168) and its active stereoisomer LY303070 are the most effective and specific antagonists (123, 187). The IC₅₀ of GYKY 53655 for AMPA receptors is ~1 µM, and it has no effect on kainate receptors at concentrations as high as 100 µM. As a result, the use of this molecule has made it possible to unmask the kainate receptors present in neurons, advancing very significantly the study of their physiological role. Other molecules related to GYKI 53655, such as GYKI 52466 and GYKI 53405, have been also used to discriminate between the two receptor subtypes (12, 39, 123). Although the affinity of these derivatives for AMPA receptors is comparable to the affinity of GYKI 53655, high concentrations can reduce the activity of kainate receptors, a fact that has limited their applicability. Similarly, SYM 2206, another diastereomer of 4-methylglutamate, has recently been used as a selective blocker of AMPA receptors (96, 128, 142), but the detailed pharmacological characterization of this molecule has not been completed (Table 3).

One final pharmacological tool that has been greatly used during the study of kainate receptors is the lectin concanavalin A. This molecule binds to a series of N-glycosylated residues present in this receptor subtype (43, 44), leading to a reduction of the fast desensitization observed upon ligand binding (71, 118, 193). It is active in native as well as in recombinant receptors, although the sensitivity of GluR7 is lower compared with the rest of the subunits (152). However, its practical use is limited because of several reasons. First, concanavalin A binds nonspecifically to sugars in cell membranes (43, 44), a fact that restricts its accessibility into more physiologically relevant preparations such as brain slices. Second, in addition to its effect on desensitization, concanavalin A seems to increase the affinity of kainate receptors by one or two orders of magnitude (78, 124), a fact that may lead to confusions about the actual biophysical and pharmacological properties of the channels. It has been proposed (92, 124) that the lectin could uncover quiescent receptors in a high-affinity, previously desensitized state that become conductive after the binding of concanavalin A. However, the actual mechanism underlying this increase in affinity remains to be elucidated. Third, it has been observed that other subtypes of glutamate receptors can also be modulated by treatment with concanavalin A. For instance, the currents through homomeric channels composed of GluR1 or NMDAR1-4a subunits are increased in the presence of the lectin (44).

	IV. PHYSIOLOGICAL ROLES OF KAINATE RECEPTORS

Top Previous Next References

The first evidence of a role for kainate receptors in synaptic function was obtained in the peripheral nervous system. Early studies had shown that peripheral dorsal root nerves from immature rats, specifically C fibers, were depolarized by kainate (1, 35). Subsequent studies demonstrated that kainate application could elicit rapidly desensitizing responses on acutely dissociated DRG cells (71), an effect markedly different from the nondesensitizing currents mediated by the kainate activation of AMPA receptors in central neurons. The realization that kainate induced a desensitizing response in recombinant kainate receptors (66, 160) and that the mRNA of the different subunits were expressed by a population of DRG cells (10, 171) led to the notion that kainate receptors were responsible for the response of the DRG neurons. This provided the first example of a physiological role for this kind of receptors, and recent studies have explored in more detail the actual molecular composition of the DRG kainate receptors. The pharmacological and biophysical profile of the channels resembles more closely the one observed for homomeric channels made of GluR5(Q) subunits: they are sensitive to LY293558, a GluR5 antagonist (13), and exhibit the same subconductance levels and desensitization rate (194) resolved for this subunit in heterologous expression systems (164, 166).

It is illustrative to remember that glutamate is released by C fibers at their termination site in the dorsal horn (75). This fact raises the possibility that the kainate receptors present at the DRG cells act as autoreceptors, modulating the release of the amino acid upon peripheral stimulation (see NOTE ADDED IN PROOF). In addition to this, it has been found that high-intensity stimulation of the primary afferent fibers evoke an excitatory postsynaptic current (EPSC) mediated by kainate receptors in the spinal cord (96), suggesting that this receptor subtype may be directly involved in nociception. Further support for this idea comes from the observation that several kainate receptor antagonists have shown analgesic activity in a variety of animal models of pain, as discussed in section IVA.

In contrast to what is observed in the DRG cells, AMPA receptor activation masks the existence of kainate receptors in essentially every central neuron. This fact made it difficult at first to unequivocally demonstrate their existence beyond the peripheral nervous system. However, a series of early observations clearly pointed to their presence throughout the brain. For instance, it was well known that kainate could act as a potent convulsive agent when administered in vivo (6, 31) and, more importantly, that the ingestion of contaminated mussels and the ensuing intoxication with domoate, a more selective agonist, also led to the appearance of seizures and produced amnesia (168). It had also been established that certain brain areas such as the CA3 and the CA1 regions of the hippocampus were particularly sensitive to the depolarizing and neurotoxic effects of kainate (25, 116, 138, 186). Thus concentrations of kainate unsuitable for activating AMPA receptors were nevertheless able to induce epileptiform activity (138). In addition, autoradiographic studies had revealed a discrete band of high-affinity kainate binding sites along stratum lucidum, the site of termination of the mossy fibers (49, 113). When the granule cells were destroyed, there was a parallel reduction in the number of binding sites (137) and a decrease in the susceptibility to the toxicity (36) and epileptogenic activity induced by kainate (53). These observations suggested that the putative kainate receptors were localized presynaptically, a hypothesis further supported by some ultrastructural evidence (23, 132) and by early studies showing that kainate could stimulate transmitter release from slices from a variety of brain regions (46, 47).

Despite this significant body of evidence, the ambiguity associated with a possible effect of kainate on AMPA receptors did not allow any definitive conclusion about the actual involvement of legitimate kainate receptors. It was only in 1993 that pure kainate receptor responses were recorded from embryonic hippocampal neurons in culture by predesensitizing AMPA channels (94). Under these conditions, it was observed that kainate elicited a small, rapidly desensitizing current that showed strong inward rectification, akin to what had been reported for homomeric GluR6(Q) receptors (40, 66). Indeed, the subsequent analysis of the kainate receptor mRNA content of individual neurons revealed that most hippocampal cells in culture expressed GluR6 and, furthermore, that there was a correlation between the GluR6(Q)/GluR6(R) ratio and the degree of rectification observed (144). Later studies have confirmed and extended these observations by demonstrating the existence of kainate receptors in several brain cell types including glial progenitors (127), cerebellar (129) and trigeminal ganglion neurons (146), and postnatal hippocampal cultures (189). It is interesting to note that, in the latter case, the properties of the kainate receptors are significantly different from those found in cultures from embryonic neurons: the currents do not desensitize fully, and no inward rectification is observed. These features resemble the characteristics of GluR5(R) ensembles and can, in principle, be explained by the age-related changes in the levels of expression of this receptor subunit (4) and by the developmental shift in the editing at the Q/R site (9).

The demonstration that pure kainate receptors did exist in nerve cells prompted the obvious question about their involvement in synaptic transmission. The advent of GYKI 53655 has made it possible to explore directly whether the stimulation of excitatory neurons is accompanied by the postsynaptic activation of kainate receptors in several brain areas. The initial observations made in hippocampal cell cultures seemed to indicate that this was not the case. Despite the fact that these neurons expressed functional kainate receptors, no EPSC could be recorded in the presence of AMPA and NMDA receptor antagonists (93). However, subsequent studies performed in brain slices have identified a collection of synapses that possess kainate receptors and have begun to reveal certain aspects of their physiological role (51).

The mossy fiber pathway, the connection between the dentate gyrus granule cells and the CA3 pyramidal neurons of the hippocampus, was the first synapse shown to employ kainate receptors (Fig. 7). The repetitive stimulation of the mossy fibers in the presence of GYKI 53655 and 2-amino-5-phosphonovalerate (APV), antagonist of choice of the NMDA receptor subtype, was associated with the appearance of a current with slow rise and decay time constants that disappeared when CNQX was added to the bath (21, 176). In addition to its slow kinetic properties (a feature common to most kainate receptor synapses hitherto described), several characteristics of this current are noteworthy. First, the need for repetitive stimulation to open the receptor channels has been interpreted by some groups as evidence for their putative extrasynaptic (and not postsynaptic) location (8, 90, 108). In this model, the glutamate accumulated after tetanic stimulation would spill away from the synapse, acting on receptors located far from the release site. Nevertheless, the observation that the kinetic features of the kainate receptor-mediated current are not affected by blocking the uptake of glutamate would seem to argue against this possibility (111). It should be kept in mind, however, that glutamate uptake is very sensitive to temperature variations (180). As the aforementioned experiments were performed at room temperature, it is conceivable that the glutamate carriers were already inactive, consequently explaining the lack of effect of the antagonists. In any case, if the kainate receptors activated by the mossy fibers are extrasynaptic, then they must be located rather close to the release sites. In favor of this idea, it has been observed that reducing the diffusion of glutamate by increasing the viscosity of the extracellular medium increases the size of the observed current (111).

View larger version (23K): [in this window] [in a new window]   Fig. 7. Kainate receptors mediate a slow postsynaptic current in the hippocampal CA3 region. A: schematic representation of the experimental arrangement. B: synaptic currents elicited on a CA3 neuron in the absence (left) or presence (right) of the AMPA receptor blocker GYKI 53655. The repetitive stimulation of the mossy fiber pathway is accompanied by a kainate receptor-mediated excitatory postsynaptic current (EPSC) resistant to GYKI 53655. This current is not observed upon repetitive stimulation of the associational-commissural fibers or of the mossy fibers in a GluR6 knockout mouse (C). [B from Castillo et al. (22). Reprinted from Neuron with permission of Elsevier Science. C from Mulle et al. (115). Reprinted by permission from Nature, Macmillan Magazines Ltd.]

Second, the activation of kainate receptors is not observed at other excitatory connections received by the CA3 neurons (21, 176). For instance, the repetitive activation of the associational-commissural pathway, axons that interconnect ipsi- and contralateral CA3 cells, generates an EPSC that is completely blocked by GYKI 53655 and APV (Fig. 7B). This result is in agreement with the early localization data (49, 113), and it indicates that the targeting of kainate receptors to the synapse is not unspecific within a given cell but that it is regulated by a mechanism yet to be discovered.

Third, the kainate receptor that mediates the synaptic current is probably heteromeric. On the one hand, the GluR5-specific antagonists LY293558 and LY294486 block the kainate-mediated response (174). On the other hand, the activation of the mossy fiber pathway in GluR6 knockout mice is not accompanied by a residual EPSC in the presence of AMPA and NMDA receptor blockers (115) (Fig. 7C). The recent observation that GluR5 and GluR6 can form heteromeric channels provides a plausible explanation to unify these disparate observations (14, 32, 122).

One final issue regarding the existence of kainate receptors at the mossy fiber pathway is their involvement in long-term potentiation of transmission at this synapse. Early findings had shown that the induction of this plastic change was resistant to CNQX and kynurenic acid (22, 74). However, the more selective kainate receptor blocker LY382884 has been recently shown to prevent the synaptic potentiation at this pathway (14). The reason for the discrepancy among these experiments is unknown. Similarly, it remains to be determined whether LY382884 exerts its action by blocking pre- or postsynaptic kainate receptors, a problem with an additional complication: the uncertainty about the locus of induction of mossy fiber long-term potentiation (99).

The search for postsynaptic kainate receptors in the pyramidal cells of CA1 has been unsuccessful (17, 50, 93). However, it has been demonstrated that the interneurons present in strata radiatum and oriens of this hippocampal region contain a population of postsynaptic kainate receptors at the connection they receive from the principal cells (30, 50) (Fig. 8). In contrast to what has been observed at the mossy fiber pathway, the activation of the interneuron receptors does not require stimulation of the afferent axons at high frequency (30). Instead, individual stimuli elicit a slow EPSC sensitive to LY293558, suggesting that the activated channels contain GluR5 subunits. The impact of these receptors in synaptic transmission is unknown, but it has been proposed that their activation is associated with an overinhibition of the CA1 pyramidal cells (30). This idea is based largely on the observation that bath application of kainate induces a large increase in the firing rate of the interneurons. However, it is hard to accommodate considering the known epileptogenic effects of kainate (6, 31). In addition, this conclusion stands in contrast to the reports on a reduction of the size of the evoked inhibitory postsynaptic current (IPSC) induced by the exogenous administration of kainate (29, 140) or by the direct activation of glutamatergic pathways (110), as discussed below. Furthermore, the fact that bath-applied kainate necessarily acts on synaptic, as well as on extrasynaptic receptors, complicates any interpretation of that finding in terms of the phasic activation of the synaptically localized channels. In this regard, an alternative proposal (50) suggests that the massive activation of synaptic kainate receptors might lead to the eventual death of the interneurons, and to the appearance of epileptic discharges concomitant to the reduction of the inhibitory drive. Although the size of the kainate-mediated EPSC is 10-fold smaller than the AMPA-mediated current, the longer duration of the former and its correspondingly longer integration time are consistent with this possibility. Furthermore, kainate application can indeed result in the neurotoxic death of the interneurons in the CA1 region (64, 162). Whether this cell death occurs on a time scale compatible with the epileptogenic effect of kainate remains to be demonstrated.

View larger version (28K): [in this window] [in a new window]   Fig. 8. Excitatory synapses on interneurons but not on CA1 pyramidal cells have a kainate receptor-mediated component. The stimulation of the Schaffer collaterals in the presence of GYKI 53655 can elicit a kainate receptor-mediated excitatory postsynaptic current on CA1 interneurons (A, C) but not on pyramidal cells (A, B). The subsequent addition of 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) is very effective in blocking the residual synaptic current (C). The currents shown in C have been enlarged in D. [Data in B-D from Frerking et al. (50).]

Postsynaptic kainate receptors have been identified in brain regions other than the hippocampus such as the cerebellum (18), the striatum (24), the spinal cord (96), and the amygdala (95). In neurons of the basolateral amygdala (95), for instance, the stimulation of external capsule afferent fibers elicits an EPSC resistant to GYKI 53655 and APV but sensitive to LY293558. These kainate receptor-mediated synaptic responses have not been observed following the stimulation of local circuits within the amygdala (e.g., the projection from the basal nucleus) (95). Thus, as is the case for the CA3 neurons, kainate receptors at the basolateral amygdala seem to be compartmentalized. In addition, the activation of the primary afferent sensory fibers produces a kainate receptor-mediated EPSC on the spinal neurons of the dorsal horn (96). As is the case for the hippocampal currents, the onset and decay time constants of the spinal EPSC are slow compared with the AMPA receptor-mediated response in the same cells. A remarkable feature of this current is that it can only be elicited upon the stimulation of the afferent axons at an intensity strong enough to activate the high-threshold A and C fibers, raising the possibility that the kainate receptor subtype may be exclusively involved in nociception at this level (96). This hypothesis has received significant support from experiments showing that the application of opiate agonists can reduce the amplitude of the kainate receptor EPSC (96), and from the observation that this receptor subtype is present at synapses on identified spinothalamic projection neurons (96). Moreover, several kainate receptor antagonists possess analgesic activity in a number of animal models of pain. For instance, SYM 2081, presumably acting as a functional antagonist, increases the latency of escape in the hot-plate and the tail-flick tests (96), whereas LY382884 decreases the frequency of paw licking induced by the subcutaneous injection of Formalin (157).

Postsynaptic kainate receptors are commonly assumed to coexist with other glutamate receptor subtypes, in particular with AMPA receptors and, in fact, there is indirect evidence that this is actually the case for some synapses (50). However, it has been clearly demonstrated that certain synaptic contacts solely contain kainate receptors. In this regard, the connection between the cones and the off-center bipolar cells of the squirrel retina is a prime example (37). The depolarization of an individual cone leads to the appearance of a fast EPSC in this class of bipolar cells (Fig. 9). This EPSC is not affected by the addition of GYKI 53655 and APV, but it is blocked by CNQX, clearly indicating its dependence on kainate receptors. In contrast to what has been observed at all other kainate receptor synapses, the kinetic properties of the EPSC resemble those described for recombinant and for somatodendritic, extrasynaptic channels (94, 124): fast activation and desensitization time constants, an essentially complete inactivation in the continuous presence of the agonist, and a slow recovery from the desensitized state. It has been proposed that these properties of the receptors have profound implications for retinal function, controlling the gain of the bipolar cell output (37). Thus, in the dark, the cones release glutamate relentlessly, but the desensitization of the kainate channels prevents the continuous depolarization and the saturation of the off-center cell. The hyperpolarization of the cones upon illumination would interrupt transmitter release, allowing for the slow recovery of the receptors from the inactive state. If darkness were reestablished before the recovery is complete, the depolarization of the bipolar cell would be smaller depending on the duration of the illumination period (37). This mechanism for the transmission of graded, phasic signals in response to tonic release constitutes an elegant counterpart by an ionotropic receptor to the well-understood action of metabotropic glutamate receptors on on-center bipolar cells (106, 117).

View larger version (42K): [in this window] [in a new window]   Fig. 9. Kainate receptors mediate synaptic transmission between cones and off-center bipolar cells in the squirrel retina. A: an off-center bipolar cell is recorded from the inner nuclear layer (INL) of the retina, and a synaptically connected cone is depolarized (B). This manipulation elicits a fast postsynaptic response in the bipolar cell that rapidly declines due to receptor desensitization. The somatic application (C) of kainate (D) or glutamate (E) to a bipolar cell is accompanied by a GYKI 53655-resistant, desensitizing response similar to the one observed after the depolarization of the cone. [From DeVries and Schwartz (37). Reprinted by permission from Nature, Macmillan Magazines Ltd.]

One final example of synapses purely containing kainate receptors is found in the cerebral cortex (Fig. 10). The stimulation of thalamocortical axons elicits an EPSC that consists of a fast AMPA receptor-dependent component and a slower one, mediated by the activation of kainate receptors (85). However, the spontaneous EPSC recorded from cortical neurons never exhibit the two components but, instead, they fall into two separate classes corresponding to the activation of each of the different receptor subtypes (85). In other words, despite the fact that a given cortical neuron can express both AMPA and kainate receptors, they never seem to colocalize at synaptic sites (Fig. 10, B and C). Furthermore, there is a dynamic switch that governs during development the number of synapses of each class (85). Thus the occurrence of the critical period for activity-dependent plastic changes in the thalamocortical pathway coincides with a reduction in the number of synapses that possess kainate receptors and with a parallel increase in the abundance of AMPA receptor-containing contacts (Fig. 10E). A similar phenomenon has been observed after the induction of long-term potentiation at this pathway (Fig. 10D), indicating that the redistribution can occur within minutes (85). These results highlight once more the necessity to identify what synaptic proteins interact with the kainate receptors to help us understand the mechanisms that control their function and their regulation. At the same time, the existence of such a rapid switch for modifying the receptor subtype present at the synaptic contact is reminiscent of what has been observed at the so-called silent synapses (103), contacts devoid of AMPA channels that only possess NMDA receptors. The functional implications of those thalamocortical kainate receptor-only synapses remain unresolved.

View larger version (26K): [in this window] [in a new window]   Fig. 10. Kainate receptors are present at thalamocortical synapses. The stimulation of thalamocortical axons is followed by an excitatory postsynaptic current (EPSC) with two phases: a fast, AMPA receptor-mediated component and a slow, GYKI 53655-resistant, kainate receptor-mediated component (A). The two phases do not coexist at the same synaptic contacts. Instead, each synapse has solely kainate (B) or AMPA receptors (C). The induction of long-term potentiation at these synapses leads to a reduction in the number of kainate receptor-containing synapses (D). A similar phenomenon is observed during postnatal development (E), where the slow phase of the EPSC disappears by postnatal day 8. [From Kidd and Isaac (85). Reprinted by permission from Nature, Macmillan Magazines Ltd.]

Another unresolved issue regarding the synaptic activation of kainate receptors is the slow kinetic properties of the response. With the exception of the channels found at the squirrel retina (37), the time constants determined for the