AMPA Receptors Bring On the Pain

Previews 577

Shen, J., and Cookson, M.R. (2004). Neuron 43, 301–304. Takai, Y., Sasaki, T., and Matozaki, T. (2001). Physiol. Rev. 81, 153–208. Wszolek, Z.K., Pfeiffer, R.F., Tsuboi, Y., Uitti, R.J., McComb, R.D., Stoessl, A.J., Strongosky, A.J., Zimprich, A., Muller-Myhsok, B., Farrer, M.J., et al. (2004). Neurology 62, 1619–1622. Zimprich, A., Biskup, S., Leitner, P., Lichtner, P., Farrer, M., Lincoln, S., Kachergus, J., Hulihan, M., Uitti, R.J., Calne, D.B., et al. (2004). Neuron 44, this issue, 601–607.

AMPA Receptors Bring On the Pain The role of Ca2ⴙ-permeable AMPA receptors in pain processing has not been extensively studied. In this issue of Neuron, Hartmann et al. show that altering the levels of these receptors has consequences for inflammatory pain hypersensitivity but not acute pain processing. In the late 1980s and early 1990s, the previously elusive receptors for the excitatory amino acid glutamate were characterized, cloned, and intensively studied. The NMDA receptor was shown to be blocked by Mg2⫹ and to be permeable to Ca2⫹. The non-NMDA receptors, named the AMPA and kainate receptors, were initially believed to show little permeability to Ca2⫹. At that time, the kainate receptor’s role in nervous system function was obscure, but the AMPA receptors were known as the key transducers of synaptically released glutamate. Their apparent lack of Ca2⫹ permeability and voltagedependent activity made them seemingly simple to understand compared to the complex NMDA receptor. However, it soon became clear that within the four genes coding subunits of the AMPA receptor family, GluR1, -2, -3, and -4, also called GluR-A, -B, -C, and -D, the presence of GluR2/B made the multisubunit AMPA receptors Ca2⫹ impermeable. In the absence of GluR2/B, AMPA receptors were permeable to Ca2⫹ and showed an inwardly rectifying dependence on membrane potential (Verdoorn et al., 1991). These Ca2⫹-permeable AMPA receptors are now known to be expressed at synapses between a variety of neurons throughout the central nervous system, including the spinal cord dorsal horn— the first stage of central somatosensory processing. Ca2⫹-permeable AMPA receptors have proven notoriously difficult to study. Because they are defined by the absence of GluR2/B, the only certain way to demonstrate their active presence on the surface of a neuron is by functional assay. Pharmacology of the receptors is also limited because the compounds that block them, including polyamines such as Joro spider toxin, are open channel blockers and so are complicated to use. In addition, these compounds are nonselective in that they also block Ca2⫹-permeable kainate receptors. In this issue of Neuron, Hartmann et al. (2004) have instead used a genetic approach to investigate the role of Ca2⫹-permeable AMPA receptors in the spinal cord dorsal horn in normal nociceptive processing versus the

altered sensory processing present in pathological pain states. The superficial dorsal horn, the main site of termination of nociceptive afferent input to the spinal cord, has previously been shown to contain a high density of Ca2⫹-permeable AMPA receptors, as identified using kainate-induced cobalt uptake in which cobalt goes through open Ca2⫹-permeable AMPA receptors (Engelman et al., 1999). Hartmann et al. have shown that, compared with wild-type littermates, mice lacking GluR1/A have reduced numbers of cobalt-positive neurons, whereas mice lacking GluR2/B have increased numbers of cobalt-positive neurons, using this same assay. In other words, mice lacking GluR1/A and GluR1/B have decreased and increased numbers of Ca2⫹-permeable AMPA receptors in the superficial dorsal horn, respectively. The AMPA receptor-mediated component of synaptic currents, recorded from the superficial dorsal horn, were reduced in the GluR1/A⫺/⫺ mice and enhanced in the GluR2/B⫺/⫺ mice. However, both groups of mice showed intact nociceptive withdrawal reflexes as assessed behaviorally. Latency to tail flick withdrawal from noxious heat was unaltered, and hindpaw withdrawal in response to noxious thermal or mechanical stimuli was similarly unaffected. Additionally, in an in vitro preparation, activity reflecting spinal cord reflexes elicited by nociceptor stimulation was not altered. This indicates that the nociceptive pathway is intact and functioning in both sets of modified animals. The behavioral hypersensitivity that develops following an inflammatory insult, however, was altered. The formalin test was used as one assay for such changes. In this test, formalin injection into the hindpaw of the animal results in an early (phase I) and late (phase II) behavioral response. The phase I response reflects C fiber activation due to the inflammatory insult, whereas the phase II response is thought to reflect central sensitization within the dorsal horn, in part because of its sensitivity to intrathecal NMDA receptor antagonists (Woolf and Salter, 2000). GluR2/B⫺/⫺ animals showed an enhanced response in phase II of the formalin test. Conversely, the GluR1/A⫺/⫺ animals showed depressed responses in phase II. This suggests that altering AMPAR composition can modulate NMDAR-mediated increases in central excitability. Consistent with this observation, the GluR1/A animals did not show elevated phosphorylation of extracellular activated MAP Kinases1/2 (ERK1/2) in lamina I following high-frequency stimulation of high-threshold C fibers, an in vitro correlate of the persistent nociceptor activation induced by an inflammatory agent like formalin, whereas GluR2/B⫺/⫺ animals showed normal elevations. These experiments clearly show that the composition of AMPA receptors expressed at synapses in the pain pathway strongly influences the generation of hypersensitivity in inflammatory pain models. The correlation between altered central sensitization and the numbers of neurons expressing Ca2⫹-permeable AMPA receptors, as shown by changes in cobalt loading, suggests that this family of receptors can shift the balance of longlasting changes in excitability within the dorsal horn. To understand the implications of this observation, however, a great deal more investigation will be needed. This is because not only can AMPA receptor type influ-

Neuron 578

ence the generation of pain states, but pain states can change AMPA receptor composition. Alterations in the levels of GluR1/A and GluR2/B subunits have been reported in chronic pain models as well as altered levels of the interacting proteins thought to be involved in the trafficking and delivery of the GluR subunits (Garry and Fleetwood-Walker, 2004). In addition, synaptic AMPAR function is dynamically regulated via phosphorylation of existing receptors and by activity-dependent delivery of synaptic AMPA receptors (Bredt and Nicoll, 2003). Indeed, this modification of AMPA receptor function and composition can result from synaptic activity mediated by the AMPA receptors themselves (Liu and CullCandy, 2000). Clearly, the role of AMPA receptors in synaptic plasticity and the development of central sensitization in the spinal cord dorsal horn is complex. This paper demonstrates, however, that it may be possible to selectively target AMPAR involvement in chronic pain states without altering the normal processing of acute nociceptive input. Carole Torsney and Amy B. MacDermott Department of Physiology and Cellular Biophysics and The Center for Neurobiology and Behavior Columbia University 630 West 168th Street New York, New York 10032 Selected Reading Bredt, D.S., and Nicoll, R.A. (2003). Neuron 40, 361–379. Engelman, H.S., Allen, T.B., and MacDermott, A.B. (1999). J. Neurosci. 19, 2081–2089. Garry, E.M., and Fleetwood-Walker, S.M. (2004). Pain 109, 210–213. Hartmann, B., Ahmadi, S., Heppenstall, P.A., Lewin, G.R., Schott, C., Borchardt, T., Seeburg, P.H., Zeilhofer, H.U., Sprengel, R., and Kuner, R. (2004). Neuron 44, this issue, 637–650. Liu, S.Q., and Cull-Candy, S.G. (2000). Nature 405, 454–458. Verdoorn, T.A., Burnashev, N., Monyer, H., Seeburg, P.H., and Sakmann, B. (1991). Science 252, 1715–1718. Woolf, C.J., and Salter, M.W. (2000). Science 288, 1765–1769.

Pruning an Axon Piece by Piece: A New Mode of Synapse Elimination

The process by which excess axons are pruned during development has remained unclear. In this issue of Neuron, Bishop et al. use time-lapse imaging and serial electron microscopy of developing neuromuscular junctions to describe a novel cellular mechanism in which retracting axon branches shed fragments rich in normal synaptic organelles. These “axosomes” are engulfed by adjacent Schwann cells and may be assimilated into the glial cytoplasm. Shedding of axo-

somes and glial engulfment may represent a widespread mechanism of synapse elimination. A common feature of nervous system development is the generation of a surplus of nerve cells as well as their connections. Subsequently, these excess cells and synaptic connections are removed in a carefully regulated fashion to generate the mature functional circuitry of the nervous system. Much of our knowledge about the elimination of synaptic connections comes from the vertebrate neuromuscular junction (NMJ), which is relatively simple and highly accessible for direct observation in vivo (Lichtman and Colman, 2000). When motor neurons first extend axons to their target muscles, each neuron connects to a far larger number of muscle fibers than in adulthood. Due to extensive overlap of this surplus innervation, each muscle fiber receives inputs from multiple axons. Even at the same synaptic site on a muscle fiber, multiple axons compete for a foothold. Then, during the early postnatal period, all but one of the inputs are eliminated such that each fiber is innervated by a single axon. Elimination of inputs is not due to motor neuron death but rather due to pruning of axonal branches, such that a motor neuron loses some synaptic connections while maintaining others. Two competing hypotheses were advanced to explain the mechanism of axonal withdrawal (Figure 1). One was that axonal branches underwent degeneration similar to the Wallerian degeneration observed in distal nerve stumps after nerve injury (Figure 1B). Features of degeneration such as fragmentation of the axon branch, aberrant mitochondria, high cytoplasmic density, and clumping of synaptic vesicles observed in electron micrographs were the main evidence cited in favor of this view (Rosenthal and Taraskevich, 1977). The second, more popular view was that axonal branches underwent distal-to-proximal retraction without degeneration and were resorbed into the parent axon (Figure 1C). The presence of “retraction bulbs,” large vesicle-rich structures at the ends of thin axonal branches, and the absence of degenerative features favored the retraction hypothesis (Bixby, 1981; Riley, 1981). The issue remained unresolved, however, in large part because the dynamics of the elimination process had not been directly observed. In a beautiful study in this issue of Neuron, Bishop et al. (2004) propose a new concept that combines some features of the above two models and provides fresh insight into the mechanisms of synapse elimination. In a superb example of using new tools to illuminate longstanding questions, the authors performed time-lapse imaging of fluorescent axons in transgenic mice during synapse elimination, complemented with serial electron microscopy, to study axon withdrawal with high temporal and spatial resolution. The double-transgenic mice that they used express YFP in all motor axons and CFP in a subset; therefore, synaptic sites undergoing input elimination (those occupied by both a YFP and a CFP axon) could readily be identified. When observing “losing” axons as they withdrew from the doubly innervated synaptic sites, the authors noticed that fluorescent remnants of axons were sometimes left behind (Figures 1D– 1F). They named these membrane bound structures “ax-