Glutamate receptors on myelinated spinal cord axons: I. GluR6 kainate receptors

The deleterious effects of glutamate excitotoxicity are well described for central nervous system gray matter. Although overactivation of glutamate receptors also contributes to axonal injury, the mechanisms are poorly understood. Our goal was to elucidate the mechanisms of kainate receptor–dependent axonal Ca2+ deregulation.

Glutamate is the main excitatory neurotransmitter in the mammalian central nervous system, playing a significant role in gray matter injury in many neurodegenerative diseases. 1 Prevalent and devastating disorders such as stroke, multiple sclerosis, and trauma to the brain and spinal cord invariably affect afferent and efferent white matter tracts, though much less is known about mechanisms of injury to myelinated white matter axons. Voltage-gated Na ϩ and Ca 2ϩ channels, together with reverse Na ϩ -Ca 2ϩ exchange, play important roles 2-4 (for review, see Stys 5 ). Perhaps counterintuitive, given the nonsynaptic nature of central nervous system white matter, are observations of functional protection of this tissue by antagonists of ionotropic glutamate receptors. ␣-amino-3-hydroxy-5-methyl-4isoxazole propionic acid (AMPA)/kainate receptor antagonists are protective both in vitro 6 -10 and in vivo, [11][12][13][14] in ischemic, traumatic, and autoimmune models of white matter injury. Conversely, activating AMPA/kainate receptors, but not N-methyl-D-aspartate (NMDA) receptors, or increasing extracellular glutamate levels by blocking glutamate transport either in vitro [15][16][17] or in vivo [17][18][19] is injurious to axons.
The precise mechanisms of injury to white matter elements induced by non-NMDA glutamate receptor activation are unknown. Both astrocytes and oligodendrocytes express AMPA and kainate receptors (for review, see Matute and colleagues 20 ), and more recently, NMDA receptors have been detected on mature oligodendrocytes, 21 their processes, 22 and even the myelin sheath. 23 These receptors are permeable to Ca 2ϩ ions; therefore, it is reasonable to conclude that receptormediated Ca 2ϩ overload is responsible for excitotoxic glial injury. 15,24,25 What is so far unexplained is the observation that central axons per se are damaged by activation of AMPA/kainate receptors 18,19 and, in turn, protected by blockers of these receptors in various injury models. 9,13,26 These latter observations raise the possibility that central myelinated axons themselves express AMPA/kainate receptors, whose overactivation results in damage to the fibers directly. Indeed, antagonists of AMPA/kainate receptors, but not NMDA receptors, were protective against spinal cord dorsal column injury, 6 -8 and bath application of AMPA, kainate, or glutamate, but not NMDA, induced irreversible reduction of compound action potential. 6,16 In this report, we tested the hypothesis that myelinated axons from rat spinal cord express functional kainate receptors capable of mediating a potentially deleterious axonal Ca 2ϩ increase. We found that GluR6containing kainate receptors reside along the internodal axolemma in "nanocomplexes" together with neuronal nitric oxide synthase (nNOS), exerting control over L-type Ca 2ϩ channels and causing Ca 2ϩ release from intraaxonal Ca 2ϩ stores. These signaling molecules are organized in a surprisingly intricate arrangement (see Fig 6) reminiscent of what is found at the postsynaptic membrane of conventional glutamatergic synapses.

Materials and Methods
All experiments were performed in accordance with institutional guidelines for the care and use of experimental animals. Additional details can be found in the supplementary material.

Immunochemistry and Immunoelectron Microscopy
Immunohistochemistry, immunoelectron microscopy, and immunochemistry were performed using standard techniques 23 (see supplemental material).

Intraaxonal Nitric Oxide Generation Promotes the Ca 2ϩ Increase
Although the earlier results support the involvement of kainate receptors in the mobilization of Ca 2ϩ , they do not prove that these receptors are necessarily axonal; indeed, the protective effects of AMPA/kainate antagonists in white matter injury was suggested to be due to protection of glial elements 33 with indirect sparing of axons (for review, see Matute and colleagues 34 ). The experiments shown in Figure 3A, relying on selective extracellular versus intraaxonal application of scavengers, strongly suggest that kainate receptors are expressed directly on axons and stimulate formation of nitric oxide (NO) within axons, which, in turn, promotes the above Ca 2ϩ release cascade. Bath application of the NO scavenger myoglobin 35 failed to prevent axoplasmic Ca 2ϩ increase (kainate ϩ myoglobin: 80 Ϯ 66%, n ϭ 27, p ϭ 0.2 vs kainate alone; SYM2081 ϩ myoglobin: 145 Ϯ 49%, n ϭ 34, p Ϸ 1). Hydroxocobalamin, another NO scavenger 36 with a much smaller molecular weight (and, therefore, more readily able to permeate small interstitial spaces between axons, but nevertheless membrane impermeable), was equally ineffective (kainate ϩ hydroxocobalamin: 90 Ϯ 71%, n ϭ 23, p ϭ 0.94 vs kainate alone). These experiments indicate that NO synthesized outside the axon did not play a role in kainate receptor-mediated Ca 2ϩ release inside axons. To explore whether intraaxonally generated NO may be important, we selectively loaded myoglobin into axons. In contrast with bath application, intraaxonal scavenger potently blocked kainate-(0 Ϯ 22%; n ϭ 22) and SYM2081-induced (16 Ϯ 33%; n ϭ 25) Ca 2ϩ responses ( p Ϸ 0). Intraaxonal hydroxocobalamin was also highly effective, as was the nitric oxide synthase inhibitor L-NAME ( p Ϸ 0). Moreover, the effect of intraaxonal NO was syner- gistic with depolarization, even in the absence of receptor activation (see Fig 3B): Neither depolarization alone (45mM K ϩ in the perfusate) nor exogenously applied NO (using the NO donor PAPA NONOate [250M]) induced an axonal Ca 2ϩ increase. However, applying the NO donor during K ϩ -induced depolarization induced a substantial axonal Ca 2ϩ increase, which was greatly reduced by either nimodipine or ryanodine.
Axonal Signaling "Nanocomplexes" Containing GluR6/7, Neuronal Nitric Oxide Synthase, and Ca v 1. 2 The previous observations suggest a close relation between axonally expressed GluR6 kainate receptors and nitric oxide synthase. Immunohistochemistry was performed to further localize these receptors and their associated signaling proteins (Fig 4). Punctate staining for GluR6/7 (using two different primary antibodies from different species) and nNOS was observed at the periphery of neurofilament-labeled axon cylinders. These clusters were often, but not invariably, colocalized. Although we did not attempt to examine the frequency of these complexes along the length of an axon, the representative micrograph in Figures 4A to C suggests that at least several clusters are present per internode. Immunoelectron microscopy localized GluR6/7 to the axolemma and to clusters beneath the axolemma. Consistent with earlier pharmacological evidence pointing to a functional interaction between kainate receptors and L-type Ca 2ϩ channels, colocalized GluR6/7 and Ca v 1.2 clusters were also observed at the surfaces of axons (see Figs 4E-G). Immunoprecipitation of dorsal column lysate with the GluR6/7 antibody yielded a single nNOS-positive band indicating a physical association between this kainate receptor and the enzyme (see Fig 4I). We further hypothesized that a PDZ-binding motif on the C terminus of GluR6 may mediate an interaction between this receptor and an adaptor protein, 37 which, in turn, may scaffold the receptor in proximity to axonal nNOS to support a functional relation. We constructed a peptide comprising the nine C-terminal residues of GluR6 (RLPGKETMA, see Materials and Methods), to interfere with such a putative interaction. When this peptide was loaded into axons, both kainate and SYM2081 Ca 2ϩ responses were almost completely blocked (kainate ϩ peptide: 12 Ϯ 28%, n ϭ 77, p ϭ 1.2 ϫ 10 Ϫ5 vs kainate alone; SYM2081 ϩ peptide: 13 Ϯ 27%, n ϭ 78, p ϭ 1.1 ϫ 10 Ϫ5 ). A sham peptide had little effect on the Ca 2ϩ increase induced by kainate (91 Ϯ 28% n ϭ 45) or SYM2081 (96 Ϯ 30%; n ϭ 42); the responses with the active compared with the sham peptides were highly significantly different ( p Ͻ 10 Ϫ9 for both agonists) (Fig 5A). Further proof of an intraaxonal localization of a GluR6-PDZ domain, which could scaffold this receptor within a signaling nanocomplex containing nNOS, was obtained by loading the synthetic interfering peptide, itself labeled with multiple fluorescent moieties, into axons. As with the fixed immunohistochemical sections, we observed occasional punctate clusters of fluorescent peptide at the periphery of fluorescein-dextran-loaded axons (see Figs  5B-D), consistent with the notion that these fibers contain discrete clusters of PDZ domains able to bind and likely cluster kainate receptors.

GluR6 Activation Causes Functional Dorsal Column Injury
Having identified such an arrangement of internodal signaling protein clusters capable of significantly increasing axonal Ca 2ϩ levels, we then explored whether such persistent increases of Ca 2ϩ had any functional implications in otherwise uninjured dorsal columns. Propagated compound action potentials were recorded electrophysiologically, and functional integrity of this white matter tract was determined by calculating the area under the digitized responses. 38 Exposure of dorsal columns to kainate (200M) or SYM2081 (100M) for 60 minutes followed by a 3-hour wash caused an irreversible reduction of mean compound action potential (CAP) area to approximately 60% of control (data not shown). Addition of the L-type Ca 2ϩ channel blocker nimodipine (10M) significantly protected against kainate-(CAP area recovery: kainate ϩ nimodipine, 93 Ϯ 17%, n ϭ 8, vs kainate alone, 68 Ϯ 10%, n ϭ 8; p ϭ 0.003, Wilcoxon rank test) and SYM2081-induced injury (SYM2081 ϩ nimodipine, 83 Ϯ 23, n ϭ 9, vs SYM2081 alone, 51 Ϯ 15, n ϭ 7; p ϭ 0.0022).

Discussion
A number of in vitro and in vivo studies have pointed to an important role for non-NMDA glutamate receptors in white matter injury, 6,8,9,16 with glial cells representing an important target given their known expression of AMPA and kainate receptors, 20 and their sensitivity to this excitotoxin. 24,39 This sensitivity to AMPA/kainate receptor activation also applies to immature oligodendrocyte precursors. 40 Glutamate is released from injured myelinated axons via reverse Na ϩdependent glutamate transport 7 and via vesicular release from unmyelinated fibers during physiological activation. 41,42 In contrast, little is known about functional glutamate receptors on central axons, though experiments indirectly suggest that such receptors may be present. 8,43 Here we show that functional kainate receptors are present on myelinated central axons, raising the dis-  tinct possibility that loss of axonal function after glutamate exposure may also be caused by direct activation of axonal receptors leading to (possibly focal) axoplasmic Ca 2ϩ deregulation. Curiously, immature premyelinated fibers are reported to suffer ischemic injury independently of glutamate receptors. 33 Contrasted with our findings in mature myelinated axons, this may indicate that myelination induces expression and clustering of axonal glutamate receptors, as it does other nodal and perinodal proteins. 44 Immunohistochemistry of dorsal column axons showed colocalized Glur6/7 and nNOS clusters sparsely distributed along axon cylinders as has been reported previously for Ca v and RyR clusters. 30 Our results are consistent with the following proposed feed-forward mechanism (Fig 6): Activation of GluR6-containing kainate receptors induces a local depolarization of the internodal axolemma, together with a small amount of Ca 2ϩ influx from a restricted periaxonal space. The local axonal Ca 2ϩ microdomain promotes NO synthesis by nNOS, and the local depolarization activates L-type Ca 2ϩ channels, thereby opening ryanodine receptors on subaxolemmal endoplasmic reticulum, culminating in a much larger Ca 2ϩ transient than would be possible solely by influx of this ion. This is consistent with previous observations of kainate receptor-mediated depolarization of central axons. 43 Our electrophysiological recordings, which showed that functional injury induced by kainate receptor stimulation was significantly reduced by blocking L-type Ca 2ϩ channels, emphasize two important points. First, given that activation of these receptors in otherwise uninjured dorsal columns results in significant functional impairment indicates that the observed Ca 2ϩ increase induced by this treatment is pathophysiologically significant and raises the distinct possibility that exposure of axons to glutamate in inflammatory or ischemic lesions, for instance, may be directly damaging to axons. Second, the significant reduction in GluR6-mediated electrophysiological injury conferred by an L-type Ca 2ϩ channel blocker further strengthens the functional connection between these receptors and Ca 2ϩ channels, as suggested by the Ca 2ϩ imaging experiments (see Fig 2) and summarized in the proposed model (see Fig 6).
The effect of NO is curious, though this modulator may function to increase the "gain" of the Ca v -RyR coupling mechanism, possibly by upregulation of RyR activity. 45 This may be necessary to ensure the fidelity of this signaling cascade, because unlike neurons and muscle cells that are not ensheathed, voltage-gated proteins such as Ca v s, which are localized to the internodal axolemma of myelinated fibers, likely experience smaller electric-field fluctuations because of the overlying myelin. Given the known promiscuous actions of NO (and its highly reactive derivative peroxynitrite), it is possible that other ion transporters, which are important for axonal impulse propagation (eg, voltagegated Na and K channels, Na-K-ATPase 46 ), may be modulated as well in response to kainate receptor/ nNOS activation. Thus, central myelinated axons contain functional complexes of several signaling proteins that are arranged in close proximity (eg, GluR6/7, nNOS, and Ca v 1.2; see Fig 4; L-type Ca 2ϩ channels and ryanodine receptors 30 ), allowing local NO production and depolarization to modulate their function. The purpose of such clusters in mature myelinated fibers is currently unknown; in developing axons, however, growth cone dynamics have been shown to be dependent on glutamate receptor activation and release of Ca 2ϩ from intraaxonal Ca 2ϩ stores, 47 indicating that ionotropic glutamate receptors and Ca 2ϩ signaling from axonal stores are functionally related from an early developmental age. Their precise physiological roles in adulthood will require further study. Scaffolding of axonal receptors and effectors such as nNOS in close proximity is reminiscent of the organization of signaling molecules at the postsynaptic density in neurons, 48,49 and it hints at highly specialized and complex machinery assembled along the internodal axolemma, where little active signaling was thought to take place.
Both glutamate-and NO-dependent toxicity are in- volved in white matter injury, and particularly in axonal damage, in crippling disorders such as multiple sclerosis. 34 The signaling clusters described in this report likely promote and amplify local Ca 2ϩ transients, and may have profound implications for axonal pathophysiology. The local release of potentially high concentrations of Ca 2ϩ through activation of such axonal "nanocomplexes" may play an important role in the genesis of focal swellings and irreversible axonal transections 50 that render the entire fiber nonfunctional.
The surprisingly complex interaction of glutamate, NO, voltage-gated Ca 2ϩ channels, and internal Ca 2ϩ stores in axons may paradoxically present unforeseen opportunities for the development of novel therapeutic strategies.