I. INTRODUCTION

Top Previous Next References

Motoneurons transform the internal actions of the brain into behavior, translating patterns of interneuronal activity into commands for skeletal muscle contraction and relaxation. Every movement, whether simple (knee-jerk reflex, postural maintenance), rhythmic (locomotion, respiration), or complex (playing the piano, hitting a baseball, speaking), is the consequence of a highly detailed and precise pattern of activity of many populations of motoneurons, convolved with the biomechanical properties of the skeletomuscle system. Although the signal processing underlying the distribution of inputs between and within motoneuron pools determines the basic features of any movement, the final arbiters of nervous system output are motoneurons. How motoneurons respond to their inputs, and how their responses are regulated, is of interest and the subject of this review.

Sherrington (1142) introduced the concept of motoneurons as the final common path, representing the penultimate link between the central nervous system (CNS) and motor behavior. Since then, motoneurons have attracted the attention of investigators studying the cellular physiology of central neurons for several reasons. 1) The function of motoneurons is well defined, i.e., causing contraction of striated muscle. 2) With the advent of intracellular recording techniques, it was possible to study the electrical properties of these large accessible neurons, which can be readily identified in physiological experiments by antidromic invasion from specific muscle nerves (141). 3) Multisensorial inputs from muscles, joints, and skin produce synaptic potentials in motoneurons, and experimental access to these pathways paved the way for studies of synaptic transmission in the CNS and of simple reflex pathways in mammals (317, 436). Numerous major reviews have been published on motoneuron physiology; among the more notable are References 49, 107, 161, 162, 475.

From Sherrington to Eccles, motoneurons were the paradigmatic neurons of the brain. Many fundamental and general properties of neurons and synaptic transmission were first identified in motoneurons, e.g., quantal release, inhibitory transmission, and the consequent conclusion that chemical neurotransmission is the principal form of interneuronal communication. In the past decade, interest in motoneurons has waned as intense investigation of other regions of the brain has led to an encyclopedic identification of neuronal properties not typically associated with motor control, e.g., long-term potentiation observed in hippocampal neurons and proposed to be a component of learning. One difficulty in contextualizing all of this information is that many of these well-characterized properties have been identified in neurons in the absence of data concerning how these neurons process signals in actual behavior. Here, motoneurons have a unique advantage since we know precisely 1) the information coding of their output signals, i.e., contraction of the innervated muscle fibers, and; 2) in many cases, their activity during complex behaviors, either indirectly by observing movements or recording muscle activity, or directly by recording from their axons or better yet their cell bodies. This provides a context for understanding and interpreting data that should be more highly valued. One goal in this review is to emphasize this perspective. We review studies, with emphasis on recent work, describing the passive and active membrane properties of spinal and cranial motoneurons, anatomical organization of their synaptic input, and the actions of different transmitter systems on the control of motoneuronal excitability (see Fig. 1). Most of these studies were done without addressing issues that were and remain the focus of many studies of motoneurons, such as the physiological relevance and characterization of 1) different motor unit types, e.g., fast fatigue resistant, fast fatigable, slow fatigue resistant; 2) the mechanisms underlying their orderly recruitment, i.e., the size principle; and 3) the synaptic inputs from multiple classes of afferent fibers. We refer the reader to several excellent reviews on these important issues (49, 107, 475, 1103). We further limit our discussion to mammalian -motoneurons, with a few notable exceptions involving motoneurons from other vertebrates, where some motoneuronal properties have been studied in more detail. There are important differences between the properties of motoneurons in newborn compared with those in adult mammals, both with respect to cellular properties and actions of transmitters; where such differences may reflect on the control of excitability, we have separated the description of the two groups.

View larger version (33K): [in this window] [in a new window]   Fig. 1. Overview of control of motoneuronal excitability. Precise timing of voluntary movements, rhythmic movements, afferent reflexes, and other motor acts is mediated primarily by excitatory and inhibitory synaptic drive to motoneurons using glutamate, GABA, and glycine. These transmitters activate ionotropic receptors generating synaptic current in motoneurons, which is convolved with intrinsic membrane properties to produce action potentials, which trigger muscle contraction. Modulatory systems, using amines, peptides, and other transmitters, act (mostly) through metabotropic receptors to modify excitability via changes in postsynaptic ion channel function and presynaptic release processes. These various modulatory systems produce changes in excitability related to the sleep-wake cycle, motivation, and exercise.

	II. MOTONEURONS

Top Previous Next References

Motoneurons are generated from progenitor cells in the ventral region of the neural tube early in embryonic development (145, 1230). The inductive signal in this process is the Sonic Hedgehog glycoprotein (SHH), which is secreted by axial mesodermal cells of the notochord (214, 319, 339, 340, 799, 1054, 1397). The extracellular matrix protein vitronectin may act as a downstream effector or a synergistic factor in SHH-induced motoneuron differentiation (803). The transcription factor MNR2 functions as a neural determination gene, and its expression initiates motoneuron differentiation (1233). Further motoneuronal differentiation requires expression of the LIM homeodomain transcription factor, Islet1 (Isl-1), since animals in which Isl-1 function has been eliminated do not generate motoneurons (980). Diversification of motoneuron subtypes in the spinal cord is controlled by the differential expression of four LIM homeodomain proteins (Isl-1, Isl-2, Lim-1, Lim-3) (32, 763, 1230, 1254, 1276). At the time of their birth, all classes of motoneurons express Isl-1 and Isl-2, but at the time of axon extension, the four LIM factors show a stereotyped expression pattern in functional subclasses of spinal motoneurons: 1) motoneurons innervating axial muscles express Isl-1, Isl-2, and Lim-3; 2) motoneurons innervating ventral limb muscles and body wall express Isl-1 and Isl-2; and 3) motoneurons innervating dorsal limb muscles express Lim-1 and Isl-2. Early-born motoneurons in the lateral motor column (LMC) at the brachial and lumbar levels synthesize a retinoid that induces late-born LMC motoneurons to migrate past the early-born LMC motoneurons to the ventrolateral spinal cord (1172); this suggests that neuronally released retinoids may coordinate subtype identity of spinal motoneurons innervating limb muscles. Genes of the LIM homeodomain class are also expressed differentially among cranial motor nuclei (1307). Their function is presently unknown; they may regulate receptors for guidance cues that direct axons selectively along distinct pathways to regions outside the spinal cord (763). Thus inductive signals from the notochord establish the identity of motoneurons, and local signals (possibly from the paraxial mesoderm) induce a differential expression pattern of LIM factors determining the motoneuronal subtype. Once the induction and functional differentiation of motoneurons has occurred, survival of developing motoneurons depends on muscle-derived factors and/or functional changes in the state of the motoneurons, i.e., from growing cells to transmitting cells (440, 936). The nature of the muscle-derived growth molecules is largely unknown, but one, hepatocyte growth factor/scatter factor, is necessary for survival of a subpopulation of limb-innervating motoneurons (1399). The molecular basis for motoneuronal innervation of specific muscles is being elucidated in model systems such as chick hindlimb muscles, zebrafish axial muscle, and Drosophila abdominal body wall muscle (cf. Ref. 325). However, the genetic determinants controlling subtype-specific development of motoneuronal morphology, intrinsic electrical properties, and CNS connectivity are largely unknown.

In the fully developed mammal, motoneuron groups are somatotopically organized (475, 502, 822, 916, 1055). Spinal cord motoneurons are in lamina IX of the ventral horn, divided into a medial and a lateral column. Motoneurons in the medial column innervate axial muscles, and those in the lateral column, present at the cervical, upper thoracic, and lumbosacral levels, innervate limb muscles. In the lateral column, motoneurons innervating distal muscles are more dorsal. In the rostrocaudal direction, motoneuron groups innervating single muscles span one to several spinal segments. Cranial, i.e., brain stem, motoneurons are not organized in a continuous column as in the spinal cord but form distinct nuclei, with an intrinsic somatotopic organization (297, 664).

The size and dendritic arborization of spinal and cranial motoneurons vary considerably. Consider, for example, cat hindlimb motoneurons. They have medium to large somata with a diameter of 30-70 µm (246, 1287, 1298, 1435) and 5-20 stem dendrites, which have a diameter of 0.5-19 µm and ramify extensively over a mean path length of ~1,200 µm, giving rise to ~150 dendritic terminations. The dendrites tend to project in the longitudinal direction (247), a phenomenon seen in many types of spinal motoneurons (178, 283, 1189).

The dendritic membrane constitutes >97% of the total membrane surface area in cat spinal motoneurons (246), with 61% of the stem dendrites and 12-33% of more distal dendrites covered by synaptic boutons (937). Consequently, the vast majority of synaptic inputs to motoneurons are dendritic, and integration of synaptic potentials is heavily influenced by the passive membrane properties of the dendrites (548, 1024). Synaptic current generated in a dendrite attenuates as it spreads electrotonically toward the soma, escaping through open channels in the membrane and charging the dendritic membrane capacitance. Determination of the geometrical features of the dendritic tree together with electrical parameters of the membrane [membrane resistance (R_m), membrane capacitance (C_m), and cytoplasmic resistivity (R_i)], provides an estimate of the attenuation of synaptic potentials. One useful parameter is the length constant, or space constant (), which is the distance over which a steady-state voltage is attenuated to 1/e (0.37) of its initial value.

Estimates of the electrotonic length of cat spinal motoneurons, based on recordings with sharp intracellular electrodes (which introduces an artifactual somatic shunt) and morphological reconstructions of the dendritic tree, are between 1.1 and 1.6  (63-65, 336, 369, 1288). Whole cell patch-clamp recordings (reducing somatic shunt artifacts) with two electrodes on the same soma from spinal cord motoneurons in juvenile rats estimates electrotonic length of uncut dendrites at 0.85  (1247). Thus a substantial fraction of a direct current in the dendrite reaches the motoneuronal soma, making the neuron relatively compact electrotonically for slowly changing currents. However, synaptic potentials have fast rise times, and the low-pass filtering properties (due to membrane charging) of dendrites distort and attenuate a synaptic signal more strongly than an applied steady-state voltage. This strong filtering property in cat motoneurons is mainly the result of a large membrane capacitance, and peak attenuations measured for fast synaptic currents in the distal dendrites range over 20- to 30-fold. A more direct approach to the measurement of dendritic attenuation is to study spinal motoneurons in culture (Fig. 2) (706, 1290). The electrotonic length of these neurons is ~0.7 ; dual recordings (one electrode on the soma and one on a dendrite) give a 1/e attenuation corresponding to ~260 µm for excitatory postsynaptic potentials (EPSP) travelling from the dendrite to the soma (706). Interestingly, EPSP travelling from the soma to the dendrite are attenuated much less (1/e attenuation of ~710 µm), a result predicted by cable theory (188). The overall picture of the passive properties of motoneurons in culture is that the soma and dendrites are roughly isopotential for slowly varying potentials with fast voltage transients arising from synaptically generated currents more strongly filtered. Thus dendritic EPSP should produce smaller and more slowly changing somatic depolarization that could bring the membrane potential above threshold for firing but with a smeared temporal relation to underlying synaptic potentials. In contrast, EPSP in the somatic region should rise sufficiently fast that if they were above threshold, an impulse would be generated in close synchrony with the incoming synaptic potential.

View larger version (10K): [in this window] [in a new window]   Fig. 2. Attenuation of synaptic potentials along a dendrite of a motoneuron in culture. Top traces: simultaneous recordings from 2 electrodes of spontaneous excitatory postsynaptic potentials (EPSP) from the soma and a dendrite. Bottom trace: difference of the 2 recordings. Note that early and late parts of the trace are dominated by somatic and dendritic EPSP, respectively. [Adapted from Larkum et al. (706).]

Spatial differences in the membrane resistivity will also affect integration of synaptic input. For example, in cat spinal motoneurons, the somatic membrane has a lower resistivity than the dendritic membrane (179, 369, 1288), perhaps due to the presence of somatic voltage-dependent K⁺ channels (179). However, this difference may instead be due to a shunt induced by the recording pipette (1247). Ongoing synaptic input will also change the resistance of the dendritic membrane and the driving force for the synaptic current, and consequently affect synaptic integration. An estimated fivefold decrease in the membrane resistivity will result if all of the excitatory boutons impinging on a spinal motoneuron release transmitter at a rate of 30 quanta/s (107). The synaptically generated shunt of the motoneuronal membrane in an intact behaving animal has yet to be measured.

The dynamic regulation of motoneuronal excitability is largely determined by voltage-gated channels, which also are the targets for several neuromodulators affecting excitability (Table 1).

All motoneurons have a fast inactivating Na⁺ current (I_Na i), which underlies Na⁺-dependent action potentials. The exact membrane distribution of inactivating Na⁺ channels (Na_i channels) is presently unknown. Because action potentials initiate in the unmyelinated initial axon segment, this region may have a higher concentration of Na_i channels (34, 66, 141, 200, 240, 861, 1114). Normally, initial segment action potentials invade the somatodendritic membrane and give rise to action potentials, with inflections in typical intracellular recordings referred to as IS and SD spikes (66, 240, 454, 870, 919). Na⁺ channels in the dendritic membrane have not been demonstrated directly in motoneurons in vivo or in acute brain slices, but there is evidence for backpropagating Na⁺-dependent spikes in dendrites of cultured spinal motoneurons (707, 766). The voltage dependence and kinetics of activation and inactivation of Na⁺ channels in motoneurons are difficult to study, because motoneurons are not ideal for voltage-clamp techniques requiring a high time resolution. Because of their complex electrotonic structure, motoneuron membrane charging is slow; reliably controlling membrane voltage, i.e., obtaining a good "space clamp," is not possible. In two-electrode voltage-clamp recordings from spinal motoneurons, the somatic I_Na i in spinal motoneurons activates and inactivates rapidly ( ~1 ms; range, 0.1-1.3 ms depending on voltage) and exhibits some steady-state inactivation at resting membrane potential (66). The somatic membrane of neonatal rat spinal motoneurons has tetrodotoxin (TTX)-sensitive 14-pS conductance Na_i channels (1068). Activation occurs between 60 and 20 mV, and inactivation kinetics are fitted by a single exponential function ( ~1-4 ms) with a half-maximal potential of 82 mV. Interestingly, recovery from inactivation is rather slow ( ~154 ms), suggesting that control of firing frequency in motoneurons is affected by recovery time from Na_i channel inactivation. However, whether these values from the neonate are comparable to values in the adult CNS is unknown.

A persistent, i.e., noninactivating, Na⁺ current (I_Na p) is present in facial, hypoglossal, and trigeminal motoneurons (205, 870, 919). I_Na p activates below spike threshold, which would accelerate subthreshold membrane depolarization to spike threshold. I_Na p may represent Na_i channels in a noninactivating gating mode, rather than a distinct type of Na⁺ channels (148, 243).

K⁺ channels are principal determinants of the subthreshold membrane behavior, action potential shape, and firing properties of motoneurons and are also important targets for neuromodulators affecting motoneuronal excitability. Several distinct types of K⁺ conductances are found in motoneurons (825).

The resting membrane potential of cat spinal motoneurons in vivo is typically between 65 and 75 mV (458, 459, 993, 1421), positive to the equilibrium potential for K⁺ (E_K). This suggests that resting membrane potential is the result of balance between outward K⁺-, inward Na⁺-, and Cl-leak conductances. The relative contribution of these leak currents in motoneurons under in vitro conditions is estimated as g_Na/g_K = 0.13 and g_Cl/g_K = 0.25 (372). Other ionic currents, such as the inward rectifier K⁺ current (I_Kir), hyperpolarization-activated inward current (I_h), transient outward K⁺ current (I_A), and Ca²⁺-activated K⁺ currents also contribute to the resting membrane potential (74, 525, 667, 701, 1261).

Inward rectifier currents, i.e., currents that reduce upon depolarization and increase with hyperpolarization, are mediated by Kir channels (481, 912, 1065). In some neurons, inward rectifiers are active at resting membrane potential, giving rise to a steady outward current, which is in balance with leak inward currents. When the membrane is relatively unperturbed, this equilibrium ensures that the membrane potential is stable near E_K. However, if the membrane is depolarized, Kir channels close (due to a voltage-dependent block by polyamines and Mg²⁺), releasing the membrane to depolarize further (anomalous rectification). Transmitters acting on the Kir channels (through second messenger systems) can have a powerful influence on neuronal excitability by reducing the stabilizing action of the Kir current. Recently, several novel inwardly rectifying K⁺ channels (Kir_2.1, Kir_2.2, Kir_2.4, GIRK1-3) have been identified in brain stem motoneurons (607, 1261). Remarkably, expression of Kir_2.4 transcripts appears restricted to brain stem motor nuclei. Transcripts appear to be absent in higher brain structures. In hypoglossal motoneurons, block of I_Kir2.2 and I_Kir2.4 by extracellular Ba²⁺ leads to depolarization and firing, suggesting that these and related conductances indeed contribute to the motoneuronal resting membrane potential.

Delayed rectifier (I_Kdr), transient outward (I_A), Ca²⁺-activated K⁺ [I_K Ca(BK), I_K Ca(SK)], and leak currents shape the membrane trajectory of action potentials and associated afterhyperpolarizations in motoneurons. The delayed rectifier is a sustained outward K⁺ current (1065) activated by depolarization, with slower activation kinetics than I_Na i (90% complete within 5 ms, Ref. 62). It contributes to the falling phase of the action potential and the fast afterhyperpolarization (fAHP). External tetraethylammonium (TEA) blocks the delayed rectifier, lengthening action potential duration and blocking fAHP in motoneurons (62, 205, 523, 870, 919, 1113, 1213, 1311). Unitary K⁺ currents of the delayed rectifier type, observed in patches from the soma of neonatal spinal motoneurons (1068), have a ~10-pS channel conductance (in normal Ringer solution), activate between 70 and 0 mV, and deactivate slowly (60 ms at 60 mV).

I_A is a transient, i.e., rapidly inactivating, outward K⁺ current activated by depolarization, that is deinactivated by hyperpolarization (481, 1065), affecting the onset and steady-state firing of motoneurons (525, 919, 1068, 1213, 1311). In trigeminal motoneurons (525), I_A activates around 55 to 60 mV (preceded by a "priming" hyperpolarization), peaks within 5 ms, inactivates with a time constant of 6-8 ms, and is partially activated at resting membrane potential (525). In spinal motoneurons of neonatal rat, I_A activates between 60 and 20 mV, quickly inactivates ( ~15-60 ms), and has a conductance of ~19 pS in normal Ringer solution (1068). In motoneurons, 4-aminopyridine (4-AP) blocks I_A, prolonging spike repolarization, and reducing fAHP (525, 919, 1213, 1311, 1425). However, it is unclear whether I_A affects the interspike interval, since interspike hyperpolarizations in motoneurons may be too small to deinactivate I_A (525). If I_A indeed is active at resting membrane potential, it could affect the onset of firing by delaying the occurrence of the first spike in response to a strong depolarizing input.

Ca²⁺-activated K⁺ currents are ubiquitous, found in all neurons. They are the result of large (BK) and small conductance (SK) K⁺ channels gated by a rise in intracellular Ca²⁺, and in the case of the BK channels, also by voltage (1065). Both I_K Ca(BK) and I_K Ca(SK) are present in motoneurons (205, 463, 523, 647, 666, 667, 826, 870, 918, 919, 1070, 1213, 1291, 1311). I_K Ca(BK), which also is called I_c, is selective for K⁺ and activates after an influx of Ca²⁺ during an action potential. In the absence of I_K Ca(BK), the falling phase of the action potential is prolonged (1213, 1291, 1311). In patches from cultured mouse motoneurons and with symmetrical K⁺ concentrations (140 mM), BK channels have a large conductance (240 pS), a sigmoidal dependence on potential, an increased open probability with increased Ca²⁺ concentration, an inactivation time constant of 40 ms at low Ca²⁺ concentration (0.5 µM), and are blocked by external TEA (825, 826). Unitary Ca²⁺-activated K⁺ currents of the BK type in soma membrane patches of rat spinal motoneurons (1070) have a conductance of 82 pS in normal Ringer solution, are activated by intracellular Ca²⁺ and depolarization, activate rapidly (within 2-3 ms with 10⁴ M Ca²⁺ internally and depolarization to 50 mV), do not inactivate in 100 µM internal Ca²⁺, and are blocked by external TEA and charybdotoxin.

I_K Ca(SK) is a Ca²⁺-activated, voltage-independent K⁺ current blocked by the bee venom apamin and is the dominant conductance underlying afterhyperpolarizations. Spike afterhyperpolarization in motoneurons is blocked by inorganic Ca²⁺ blockers, intracellular injection of Ca²⁺ chelators, and notably by apamin (7, 205, 523, 647, 1311, 1426), suggesting that motoneurons express SK channels and that they play a critical role in the repetitive firing behavior of motoneurons. In mouse motoneurons in culture, unitary currents from SK channels have a ~18-pS conductance, a 3.5-ms mean open time, and show no voltage dependency (825).

Na⁺-activated K⁺ channels (I_K Na) are found in membrane patches from spinal motoneurons (1069). I_K Na does not appear to contribute to single action potentials but gives rise to the postdischarge hyperpolarization that follows trains of action potentials, due to accumulation of internal Na⁺.

A hyperpolarization-activated current (I_h, or I_Q in some studies), which is a mixed cationic current carried by Na⁺ and K⁺, is found in motoneurons (7, 62, 74, 205, 523, 577, 870, 919, 957, 1214, 1313). I_h has relatively slow kinetics ( ~100-400 ms, Ref. 74) and has a reversal potential positive to resting membrane potential (approximately 40 mV, Refs. 74, 1214), i.e., the net current is inward both at rest and at hyperpolarized potentials. Consequently, I_h opposes membrane hyperpolarizations, such as would be produced by inhibitory synaptic input. Activation-deactivation of I_h following hyperpolarizations gives rise to a postinhibitory rebound (PIR) and can lead to rebound bursts of action potentials. I_h may be partially active at rest, thereby contributing to the resting membrane potential (74, 701). A steady activation at rest also means that I_h affects motoneuron responses to depolarizing inputs. During depolarization, I_h deactivation will increase neuronal input resistance. When the depolarizing current or synaptic input is relieved, steady outward currents repolarize the membrane, generating an afterhyperpolarization because of the increased input resistance. The afterhyperpolarization then activates I_h, and the membrane returns to rest. (Fig. 3; Refs. 870, 957). Thus I_h seems to stabilize the membrane potential around rest and also underlies rebound depolarizations and hyperpolarizations. Spinal motoneurons in the newborn rat receive phasic excitatory and inhibitory synaptic input during neurochemically induced fictive locomotion (486), and I_h may play a role in generating rebound burst firing following phasic inhibitory synaptic input (98).

View larger version (33K): [in this window] [in a new window]   Fig. 3. Supra- and subthreshold membrane behavior of motoneurons. A: ionic currents underlying the action potential waveform. B: ionic currents underlying subthreshold membrane behavior, in this case, elicited by a short-lasting depolarizing/hyperpolarizing square current pulse. C: different phases of adaptation during repetitive firing and postdischarge hyperpolarization after a long-lasting current pulse. Unless noted, currents are activated at times indicated. For definitions, see Table 1 and section IIC.

At least six types of voltage-gated Ca²⁺ channels (L, N, P, Q, R, and T types) are expressed in CNS neurons (1030, 1275). The pharmacological and single-channel properties of motoneuronal Ca²⁺ channel subtypes have been worked out in greatest detail in neonatal hypoglossal motoneurons (1291, 1293, 1310). Three types of high-voltage-activated (HVA; L, N, and P type) and one type of low-voltage-activated (LVA; T type) Ca²⁺ channel types are present (1291), with single-channel conductances (with 110 mM Ba²⁺ as charge carrier) of 28 pS (L type), 14 pS (N type), 20 pS (P type), and 7 pS (T type). L- and P-type channels do not inactivate, whereas T- and N-type channels do ( ~20 and 58 ms, respectively). Whether adult hypoglossal motoneurons express all of these types of Ca²⁺ channels is not known. Neonatal facial motoneurons also have HVA and LVA Ca²⁺ channel types in their somatic membrane (995). However, the HVA P-type channel is absent in the somatodendritic membrane, and a novel type (R_slow) carries a major component of the total HVA Ca²⁺ current (995). In this study, single-cell RT-PCR was used to detect mRNA for the _1A-subunit of HVA Ca²⁺ channel subtypes, along with other ₁-subunit types. Because _1A-subunits are thought to combine to form P/Q-type Ca²⁺ channels, the absence of P-type channels in the soma suggests that P-type channels may be present elsewhere in the membrane of facial motoneurons, most likely in axon terminals (995). LVA and HVA Ca²⁺ channel types are also present in spinal motoneurons (524, 881). Ca²⁺ conductances affect the falling phase of the action potential, spike afterdepolarization (ADP), and afterhyperpolarization (AHP), in the latter case via I_K Ca (463, 523, 524, 647, 870, 1291, 1310, 1311). The Ca²⁺ channel subtypes underlying ADP, which is most prominent in neonatal animals, and AHP can be distinguished pharmacologically and result from both LVA and HVA as well as HVA Ca²⁺ channels, respectively (647, 1291, 1311). The L-type Ca²⁺ channels play a central role in some motoneurons that are capable of generating plateau potentials (521, 527). In the absence of modulatory transmitters, inward current flowing through L-type Ca²⁺ channels is curtailed by outward currents (524), effectively generating a small inward rectification just subthreshold to spike initiation (870, 1112, 1111). Perturbations of the balance of inward and outward currents, e.g., in form of transmitter-induced reduction in outward K⁺ currents, can lead to plateau potentials and bistable membrane behavior by uncovering inward currents through L channels (121, 285, 521, 1200). In some specialized motoneurons, a separate mechanism operates to produce plateau potentials. Motoneurons in the rostral compact formation of the nucleus ambiguus, which innervate esophageal muscles, have a Ca²⁺-activated Na⁺ current (I_Na Ca) that produces prolonged plateaulike firing in response to brief afferent input or current injection. I_Na Ca resembles the Ca²⁺-activated nonspecific cationic current (I_CAN) found in other CNS neurons (962). Spinal motoneurons in the turtle also have an I_CAN, but in these motoneurons the current is not involved in generation of plateau potentials (970).

The spatial distribution of active conductances over the somatodendritic membrane is obviously important in determining the motoneuronal response to inputs. In turtle spinal motoneurons, Ca²⁺ conductances are present in dendrites (522), which may have important consequences for the transfer of synaptic input (1162), and in the generation and modulation of plateau potentials (521, 522, 524, 1200). In rat spinal motoneurons, L-type Ca²⁺ channels are present in the somatic and proximal dendritic membrane and N-type Ca²⁺ channels in both the dendritic and somatic membrane (1351).

The relationship between synaptic input to a motoneuron pool and the resulting muscle force, i.e., the motor pool input-output function, is generally described by a sigmoid curve (107). The initial upslope of the curve results from the orderly recruitment of motor units with increasing size, axonal conduction velocity, and fatigability, i.e., the size principle (107, 475). Rate modulation of the firing of motoneurons underlies further increases, with some contribution of continued recruitment. One of the main goals in the study of motoneuronal excitability is to understand how integration of synaptic input produces this input-output relationship, and how it is affected by neuromodulators. In broad terms, transformation of synaptic input to generation of impulses depends on the following: 1) motor unit type (types S, FR, FI, FF), 2) location of synaptic terminals, 3) character and distribution of active and passive membrane properties, and 4) effects of neuromodulators on synaptic transmission and on repetitive firing behavior.

In section IV, we discuss the mechanisms by which synaptic inputs affect the input-output relationship of motoneurons. In this section, we describe the basic firing properties of motoneurons.

Motoneurons fire repetitively in response to sustained suprathreshold excitatory input. This behavior has been studied mainly by injection of current pulses through an intracellular electrode. There are distinct phases of spike-frequency adaptation to such pulses, and a nonlinear rise in steady-state firing frequency with increasing current (Fig. 3; Refs. 437, 622, 625). At the onset of a long current pulse, an initial doublet [two spikes at short interval (<20 ms)] is often seen, which is part of the initial adaptation phase (0.5-1 s) during which the firing frequency drops sharply. A second phase of adaptation follows, which in some motoneurons (1087) is divided into an early (~2-s duration) and late phase (approaching steady state for long pulses). At the end of the current pulse, there is a postdischarge hyperpolarization followed by a return to resting potential.

Initial doublets, with instantaneous frequencies of 50-300 Hz, are seen in spinal motoneurons during endogenous motor behavior (488, 639, 673, 1417), suggesting that they are not an experimental artifact. Initial doublets may permit motoneurons to generate extra force at the onset of a contraction (476, 838), perhaps to overcome inertia, but the phenomenon is not necessarily correlated with a physiological need for strong contractions (639, 673). An increase in the magnitude and duration of spike AHP contributes to the initial adaptation (48, 62, 1312), dependent on Ca²⁺ entry (activating I_K Ca) during the first few spikes (1088, 1301). However, blockade of Ca²⁺ influx does not entirely abolish initial adaptation. Other processes may contribute, such as deactivation of I_h at the onset of the depolarization, and changes in the threshold for action potential generation in the initial segment (1114). Late adaptation is also not abolished by Ca²⁺ channel blockers; in fact, it increases (1088, 1311), indicating that a number of Ca²⁺-independent mechanisms contribute to late adaptation. During repetitive firing, spike duration lengthens, and spike amplitude and rate of rise decrease (1312); this suggests that conductances that shape impulses change during maintained firing, leading to late adaptation. Late adaptation may be produced by a progressive increase/decrease in an unidentified outward/inward current (1088). Postdischarge hyperpolarization is blocked by ouabain, suggesting that it is due to an outward current generated by a Na⁺-K⁺ pump driven by local accumulation of Na⁺ (1088).

The relationship between steady-state firing rate and injected current (F-I relationship) increases in a linear fashion at low firing rates (primary range), then enters a steeper linear phase (secondary range) at higher firing rates (47, 623, 1109). In cat lumbosacral motoneurons, the secondary range starts when steady-state firing reaches ~50 spikes/s, with a slope 2-6 times that of the primary range (623). Many motoneurons lack the secondary range at steady state, with a linear F-I relationship over the entire firing range (573, 859, 870, 919). A persistent inward current (I_i, Ca²⁺ mediated) may underlie the secondary range (1003, 1114). The F-I relationship is highly dependent on the magnitude and duration of the spike AHP, since repetitive firing rate is directly related to the AHP duration (624). An effective way of modulating motoneuronal excitability is to modulate the magnitude of the AHP. For example, 5-hydroxytryptamine (5-HT) reduces the spike AHP (see sect. IVE) in cranial motoneurons, which leads to a dramatic increase in the slope of the F-I relationship (91, 526).

A particularly intriguing firing property of some motoneurons is the generation of repetitive firing that outlasts the period of excitatory input (239, 244, 326, 518-520, 527, 634, 1110). Plateau potentials underlie this behavior, which can be elicited in motoneurons by either short trains of excitatory synaptic input or current injection. Short-lasting inhibitory input can turn off the plateau potential, leading to bistable firing or membrane bistability under some circumstances. In most motoneurons, plateau potentials are not an endogenous membrane behavior but a latent property uncovered by activation of monoaminergic receptors (632, 634), and under modulatory control by other neurotransmitters (285, 1200). Thus motoneurons in spinalized cats (which are deprived of input from brain stem monoaminergic neurons) will exhibit plateau potentials when 5-HT and norepinephrine (NE) precursors are given intravenously (239, 244, 519), but not otherwise. A persistent Ca²⁺ current, possibly located in the dendrites, is the proposed ionic mechanism for generation of these plateau potentials in motoneurons (521, 522, 527, 720; see sect. IVE). During fictive locomotion (486, 1104) and in tonic muscle contractions associated with postural control (327, 633, 635), plateaulike firing is present in some motoneurons (486, 1104). Long-lasting plateau potentials are preferentially found in motoneurons with low thresholds for spike initiation and slow axonal conduction velocity, a hallmark of motoneurons of the S and FR type, which underlie most postural tasks (721, 722). The threshold for somatic activation (via current injection) of plateau potentials in cat spinal motoneurons is lowered by tonic excitatory afferent input (88). Because the plateau threshold can be lowered to the recruitment level of these motoneurons, plateau potentials under normal circumstances could play a role in amplifying the recruitment step rather than generating bistable behavior (88). An alternative view holds that plateau potentials serve to reduce the need for steady ongoing synaptic drive during maintained postural muscle contraction, through generation of self-sustained firing after transient synaptic input (634). Plateau generation in cat spinal motoneurons exhibits the phenomenon of "warm up," i.e., a progressive lowering of threshold for plateau activation with repeated activation (3- to 6-s intervals) (89). Plateau warm up may represent a form of short-term plasticity in motoneurons that ensures an increased motoneuronal output during sustained motor acts such as repetitive movements, e.g., locomotion. Specialized motoneurons in the compact formation of the nucleus ambiguus (innervating the esophagus) show plateau potentials in response to short-lasting synaptic input or current injection (1041). This plateau potential is carried by an I_CAN-like current and may generate prolonged spike activity in the ambiguus motoneurons during swallowing.

	III. ORGANIZATION OF SYNAPTIC INPUT TO MOTONEURONS

Top Previous Next References

Adult cat spinal motoneurons receive ~50,000-140,000 synaptic boutons (937), with >93% of the receptive membrane area in the dendrites. In L₇ cat motoneurons, 61% of this space is covered by synaptic boutons. GABA/glycine-immunoreactive boutons dominate the stem dendrites, covering 69% of the membrane; glutamate-like immunoreactive terminals comprise 18% (ratio ~4). In more distal dendrites, the GABA/glycine-to-glutamate ratio falls to 1.5. About 6% of the dendritic boutons are not immunoreactive for GABA, glycine, or glutamate. Presumably these boutons contain other transmitters (937).

The origins of these synaptic inputs are of considerable interest, since they encode the functional significance of the incoming signals. In the following section we briefly summarize the anatomical organization of synaptic inputs to spinal and cranial motor nuclei. The majority of the cited studies are based on tracing techniques combined with immunohistochemical detection of putative transmitters. We do not attempt to give a complete description of all the known pathways; rather, we emphasize the major anatomical pathways and the organizational principles (Table 2, Fig. 4).

View larger version (45K): [in this window] [in a new window]   Fig. 4. Anatomical organization of synaptic input to motoneurons. Main synaptic input to both cranial and spinal motoneurons comes from premotor and interneurons located close to the brain stem and spinal motoneuron pools; the few notable exceptions include direct corticospinal and rubrospinal inputs to motoneurons controlling the distal musculature, especially the digits, vestibulospinal projections to postural muscles, and bulbospinal projections transmitting inspiratory drive to phrenic motoneurons. Several cranial and spinal central pattern generators (CPG) are embedded in these premotor systems. The local premotor and interneurons also form the main gateway for relaying and integrating multisensorial afferent input from muscle, joints, skin, and descending synaptic information from forebrain, cerebellum, some brain stem nuclei, and the raphe, locus coeruleus, and other pontine/brain stem regions. Long projections from brain stem and pontine nuclei, both from diffusely projecting premotor groups, e.g., raphe and locus coeruleus, and from premotor groups involved in specialized motor tasks, e.g., respiration, equilibrium, posture, project directly to motoneurons, and to the local premotor and interneurons. Glutamate, GABA, and glycine are the principal transmitters of local premotor and interneurons but are also used in certain brain stem/pontine projections. 5-Hydroxytryptamine (5-HT), norepinephrine (NE), thyrotropin-releasing hormone (TRH), substance P (SP), and a host of other peptides are the main transmitters in the projections originating in brain stem/pontine nuclei, subserving modulatory functions in control of motoneuronal excitability. Symbols (solid circle, small open circle, large open circle, and fork shape) indicate different anatomical projection systems. Mns, motoneurons.

Monosynaptic input to spinal motoneurons from sources outside the neuraxis originate exclusively from muscle spindle Ia and group II afferents (147, 502, 543, 830, 1184). The Ia projection likely uses glutamate as a transmitter (939). Propriospinal neurons provide the major synaptic input to spinal motoneurons (502). Labeling of spinal premotor neurons has been achieved by transneuronal transport of wheat germ agglutinin, virus, or retrograde labeling following discrete injection of tracers in motor columns. The pattern of premotor neuron labeling varies considerably depending on the motor group studied (21, 227, 513, 515, 566, 1012, 1264). Some spinal motoneurons receive synaptic input via recurrent axon collaterals from other motoneurons innervating the same or synergistic muscles (248, 249). Recurrent inhibition is mediated by Renshaw neurons that send inhibitory (GABAergic and glycinergic) projections to motoneurons innervating mainly proximal muscles (409, 1099, 1370). Interneurons and propriospinal neurons from most of the spinal cord laminae relay afferent signals from muscles, joints, and skin (49, 475, 563, 869, 1064), as well as segmental (22, 534, 1236) or supraspinal input to spinal motoneurons. They also generate coordinated, rhythmic patterns of activity such as locomotion and scratching that are ultimately transmitted to motoneurons (533, 1240, 1355); only a few studies have identified rhythmically active propriospinal interneurons with direct connections to spinal motoneurons (533, 1240).

Spinal motoneurons receive extensive projections from the brain stem. The raphe pallidus and obscurus contain premotor neurons that project directly to spinal motoneurons via the lateral and ventral funiculi (69, 508, 802). There are monosynaptic projections from the raphe pallidus to deltoid motoneurons (23) and medullary raphe nuclei to phrenic motoneurons (296). Raphe-spinal projections send off collaterals to the ventral horn of the spinal cord (529) and to the intermediolateral cell column (where the other class of nervous system efferents, preganglionic neurons, are located in the spinal cord), suggesting coordination of autonomic and somatic motor activity through common premotor neurons (20). Raphe premotor neurons are either serotonergic (265) or nonserotonergic (1161). Substance P (134, 209, 495), thyrotropin-releasing hormone (TRH) (134, 471), enkephalin-like peptides (134, 397, 497, 531, 723), cholecystokinin (792), neurokinin A (907), galanin (41), and glutamate or aspartate (908) colocalize with 5-HT (based on immunohistochemistry) in neuronal cell bodies in the caudal raphe and medial reticular formation, as well as in fibers and terminals in the ventral horn of the spinal cord (39, 41, 42, 133, 575, 576, 907, 908, 1034, 1237, 1297, 1348). Peptidergic immunoreactivity largely disappears from the inputs to motoneurons after destruction of serotonergic neurons with neurotoxins such as 5,6-dihydroxytryptamine (426, 576).

Noradrenergic premotor neurons projecting to spinal motoneurons are in the locus coeruleus, the subcoeruleus, the medial and lateral parabrachial nuclei, and the Kölliker-Fuse nuclei (222, 232, 503, 507, 585, 925, 1011, 1352, 1353). Met-enkephalin or glutamate is colocalized with NE in spinally projecting locus coeruleus neurons (395, 398, 749). Intraventricular injection of 6-hydroxydopamine reduces the ventral horn content of NE and the number of noradrenergic fibers by ~85%, suggesting that the brain stem premotor neurons are the main source of noradrenergic input to spinal motoneurons (715).

Spinal motoneurons also receive direct brain stem inputs from the retroambiguus nucleus (501, 1302), ventromedial medulla (GABAergic and glycinergic neurons; Refs. 504, 506), ventrolateral medulla (331, 354, 1047, 1248), and the nucleus tractus solitarius (NTS; Refs. 331, 874). On the basis of ultrastructural analysis of synapses in the ventral horn, a glutamatergic projection to spinal motoneurons originating in the ventromedial medulla may exist, but the location of the medullary glutamatergic premotor neurons is unknown (505). The vestibulospinal system, including neurons in the medial and lateral vestibular nuclei, projects to head and neck motoneurons and lumbosacral motoneurons (299, 448, 545, 1057, 1368).

Most corticospinal projections to motoneurons are indirect, typically via interneurons in the intermediate zone of the spinal cord. Direct corticospinal projections to some motoneuron groups innervating distal musculature and perhaps functionally involved in control of fine movement, are present in primates and to a limited extent in rats (125, 146, 252, 714, 741). Finally, the red nucleus projects monosynaptically to spinal motoneurons innervating distal muscles (391, 392, 500). Several other premotor neuron systems likely project to spinal motoneurons, since transmitter-like substances are found in fibers/boutons in spinal motor nuclei, and several transmitter receptors and physiological actions of transmitters have been demonstrated in motoneurons. The anatomical location of the premotor neurons using a particular transmitter remains unknown in most cases (Table 2).

Proprioceptive information from the muscles of mastication reaches the trigeminal and hypoglossal motor nuclei via the trigeminal mesencephalic nucleus (1015). Afferent information from other peripheral sensors enters the brain stem through vagal, glossopharyngeal, and accessory nerves and is conveyed to the trigeminal, facial, and hypoglossal nuclei via premotor neurons in the nucleus of the solitary tract (83, 122, 759, 1235, 1263). A third major sensory afferent input to orofacial nuclei is from the spinal trigeminal complex (122, 482, 544, 737, 1263). The largest concentration of premotor neurons to the orofacial motor nuclei is in the medullary and pontine reticular formation adjacent to the motor nuclei themselves. Thus hypoglossal premotor neurons are ventrolateral and dorsolateral in the medullary reticular formation (122, 297, 1263), and the majority of trigeminal and facial premotor neurons are in the pontomedullary and parvicellular reticular formations (482, 535, 850, 1263). Some of these premotor neurons are GABAergic, glycinergic, or glutamatergic (738, 739, 1210, 1277, 1278). In addition to these regions, a smaller number of premotor neurons to trigeminal, facial, and hypoglossal nuclei are located in 1) pontine nuclei (Kölliker-Fuse; parabrachial nucleus; and trigeminal, facial, and hypoglossal nuclei); 2) periaqueductal gray of the midbrain (facial and hypoglossal nuclei); 3) periambigual region (hypoglossal and facial nuclei); 4) vestibular nuclei (facial nucleus); 5) gigantocellular reticular nucleus (all 3 nuclei); and 6) paralemniscal zone in the lateral midbrain and external cuneate nucleus (facial nucleus) (175, 297, 474, 482, 544, 737-739, 935, 1209, 1263, 1404).

Noradrenergic input to the hypoglossal nucleus comes from neurons in three pontine regions, i.e., nucleus subcoeruleus, A7 and A5 cell groups (12, 14). The facial nucleus receives input from noradrenergic neurons in the A5 cell group (451) and trigeminal motor nucleus from the A7 cell group (451). This differential distribution of noradrenergic input to brain stem (and spinal cord) nuclei suggests that noradrenergic neurons can be divided into subgroups that differ in their connections and functional capacities (452).

The raphe pallidus, obscurus, and magnus are the main regions containing 5-HT-positive neurons projecting to the trigeminal, hypoglossal, and facial nucleus (376, 740, 789, 790). The raphe nuclei also contain premotor neurons positive for several neuropeptides. Substance P-like immunoreactive neurons in the caudal raphe project to the trigeminal, hypoglossal, and facial motor nucleus, and Met-enkephalin-like immunoreactive premotor neurons are in the caudal raphe and medial reticular formation (375, 376, 477).

Some premotor neuron groups projecting to the orofacial nuclei are involved in dedicated motor tasks and thus have more restricted projection patterns. The central subnucleus of the solitary tract contains the pattern generator for swallowing and conveys direct synaptic information to hypoglossal motoneurons and motoneurons forming the compact formation of the ambiguus nucleus (30, 67, 467). The Edinger-Westphal nucleus projects to the facial nucleus, forming part of the circuit mediating the corneal blink reflex (645).

The nucleus ambiguus contains esophageal, pharyngeal, and laryngeal motoneurons (106). Premotor neurons projecting to the ambiguus nucleus arise from the nucleus of the solitary tract (including the swallowing-related central subnucleus), zona intermedialis reticularis parvicellularis, pontine nuclei (Kölliker-Fuse, parabrachial nucleus), vestibular nuclei, periambigual regions, paraventricular hypothalamic nucleus, external cuneate nucleus, area postrema, and periaqueductal gray (67, 467, 591, 698, 924, 998, 1130, 1144, 1427, 1428).

Surprisingly, few projections have been demonstrated from the cortex to orofacial motor nuclei (1098), emphasizing the general scheme that voluntary motor commands to motoneurons likely pass through various groups of brain stem and/or spinal premotor neurons.

Eye muscles are innervated by motoneurons in the oculomotor, abducens, and trochlear motor nuclei. Several neuronal circuits in the brain stem and midbrain are dedicated to the coordination of these motor groups, and the organization of afferent input is consequently complex. Reticular formation premotor neurons projecting to the oculomotor nucleus are in the medial midbrain reticular formation (891), reticular formation of the mesodiencephalic junction (896, 1148), and the dorsal paragigantocellular reticular nucleus (191). Premotor neurons to the abducens and trochlear motor nuclei are also in the pontine reticular formation (699, 1115, 1344, 1420). All three motor nuclei receive input from the vestibular nuclei as part of the vestibulo-ocular reflex (225, 341, 434, 685, 699, 943, 1051, 1283). Some of these premotor neurons are GABAergic (superior vestibular nucleus) or cholinergic (the medial vestibular nucleus; Refs. 191, 685, 1346). Coordination between motoneurons in the oculomotor and abducens nuclei is partly mediated by GABAergic internuclear neurons projecting contralaterally between the nuclei (284, 817, 1346). The nucleus prepositus hypoglossi contains premotor neurons (possibly glycinergic) that project to all three oculomotor nuclei (341, 684, 699, 816, 1181, 1344). The oculomotor nucleus and the trochlear nucleus receive afferents from the rostral interstitial nucleus of the medial longitudinal fasciculus (GABAergic and glutamatergic; Refs. 1180, 1335) and interstitial nucleus of Cajal (654). A small number of neurons in the locus coeruleus, trigeminal sensory complex and olivary pretectal nucleus project to the oculomotor nucleus (191, 455, 646).

Several other premotor neuron systems likely project to cranial motor nuclei (see Table 2).

There are several common principles underlying the organization of synaptic input to spinal and cranial motoneurons (Fig. 4).

1) Multisensorial afferent input is conveyed via sensory nuclei in the brain stem and spinal gray.

2) Premotor neurons located close to the motoneuron groups in the reticular formation or spinal gray provide the major synaptic input to motoneurons. One notable exception is the phrenic motoneuron pool that receives inspiratory drive from premotor neurons in the medulla. These local premotor neurons form the main gateway for relaying and integrating multisensorial afferent input and descending synaptic information. Several spinal and brain stem central pattern generators (CPG) (locomotion, scratching, and others) are embedded in this premotor system.

3) Long projections from brain stem and pontine nuclei, both from diffusely projecting premotor groups (e.g., raphe, locus coeruleus) and from premotor groups involved in specialized motor tasks (e.g., respiration, eye movements, equilibrium), converge on brain stem and spinal premotor/interneurons as well as motoneurons.

4) Some motoneurons controlling distal limb muscles receive direct cortico- and rubrospinal synaptic input.

	IV. TRANSMITTER MODULATION OF MOTONEURONAL EXCITABILITY

Top Previous Next References

Motoneuronal inputs are affected by the presynaptic release of various transmitters (amino acids, amines, peptides) acting on postsynaptic receptors; several of these transmitters can also affect signaling by actions at presynaptic receptors. The integrative effect of these transmitters acting on their receptors, in concert with the intrinsic motoneuron properties, determine the generation of the efferent signals, action potentials propagated along peripheral motor nerve fibers and recurrent collaterals. The integration is complex. Actions at ionotropic receptors induce (or reduce) localized current flows that spread according to membrane properties and cellular morphology. Actions at metabotropic receptors initiate second messenger cascades that have myriad effects, including altering channel or receptor function. Many of these actions are convergent, that is, they ultimately act via the same signal transduction mechanisms, or affect the same target, such as a specific type of channel.

In this section, we review the principal neurotransmitter systems affecting motoneuronal excitability. In each section, we discuss the important ligands, the associated receptors, the effects on neuronal properties affecting excitability, the signal transduction pathways, and, where known, the function.

Glutamate is the principal fast excitatory neurotransmitter in the CNS (226, 854, 890). The potent excitatory action of glutamate on CNS neurons was first reported in 1960 for sensory and motoneurons of cat spinal cord (262). In this section we briefly review 1) glutamate receptor ligands; 2) the molecular biology of glutamate receptors; 3) the distribution of glutamate receptor subtypes/subunits on different motoneuron pools; 4) functional implications of the structural diversity of glutamate receptors as it relates to motoneuron excitability in adult, during development, and in disease; 5) pre- and postsynaptic actions of glutamate; and 6) role of glutamate receptors in synaptic integration, production of oscillatory behavior and synaptic transmission.

A major difficulty with establishing a role for glutamate and associated excitatory amino acids (EAA; L-aspartate, L-homocysteate; Ref. 342) in synaptic signaling is that they are also involved in metabolic processes. Thus determining whether their presence, synthesis, release, and transport underlies a metabolic or signaling function is difficult (469). The relative roles of the different EAA remain unclear. Limitations of the various techniques applied to the problem are reviewed elsewhere (469). Data are most consistent with L-glutamate as the main EAA at motoneuron synapses. Circumstantial evidence includes presence of high-affinity uptake mechanisms and high glutamine levels (456, 1151) in neurons and terminals that synapse on motoneurons (144, 808); detection of increased glutamate levels in perfusate from in vitro preparations following synaptic activation or glutamate uptake inhibition in rat and frog spinal cord (443, 614, 1223); potentiation of endogenous activity by EAA uptake inhibitors (142, 442, 818); and immunohistochemical detection of glutamate in boutons synapsing on motoneurons (878, 937, 1207). Furthermore, group Ia primary afferent boutons synapsing on retrogradely labeled motoneurons are enriched in glutamate-like immunoreactivity (939).

Three ionotropic glutamate receptor subtypes, each assembled from an unknown combination of receptor subunits (most likely 5, but see Ref. 1059), are classified on the basis of pharmacological and functional properties as follows: 1) -amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) [comprising glutamate receptor (GluR) 1-4 subunits]; 2) kainate (KA: high affinity; comprising GluR5-7 and KA1, KA2 subunits); and 3) N-methyl-D-aspartate (NMDA) receptors (comprising NMDAR1, NMDAR2A-D and NMDAR3A subunits). Two orphan subunits, ₁ and ₂, have also been identified (58, 100, 294, 499, 837, 853, 880, 1100, 1118, 1119).

The functional complexity of glutamate receptor-mediated synaptic signaling is conferred by variation in receptor subunit composition and further enhanced by posttranscriptional processing of gene transcripts through alternative splicing and RNA editing. All AMPA receptor subunits undergo alternative splicing of COOH-terminal sequences and flip and flop sequences near the M4 transmembrane domain. GluR2 subunit mRNA are edited at the Q/R site and GluR2, GluR3 and GluR4 subunit mRNA are edited at the R/G sites. Kainate subunit mRNA, GluR5 and GluR6, are edited at the Q/R site. GluR6 mRNA is also edited at I/V and Y/C sites. The NMDAR1 subunit exists as eight possible splice variants (NMDAR1a-4a and NMDAR1b-4b) generated through alternative splicing of one cassette in the NH₂-terminal region and the individual or combined deletion of two cassettes in the COOH-terminal region (711, 887, 1194).

AMPA, NMDA, and kainate receptors have been identified on motoneurons via receptor autoradiography (RAR), immunohistochemistry (ICC), and in situ hybridization (ISH). However, as receptor subunit composition and posttranscriptional modification contribute to the pharmacological and physiological properties of the GluR subtypes, we will focus on ICC and ISH studies. These data are summarized in Tables 3-5 and provide specific information on receptor subunit/splice variant expression within different motor nuclei and therefore may reveal a structural/molecular basis for motoneuron pool-specific differences in physiological properties and susceptibility to excitotoxicity/motoneuron disease. The functional implications are further explored in section IVA4.

A) SPINAL MOTOR NUCLEI. Low levels of AMPA, kainate, and NMDA binding sites are seen in lamina IX of the spinal cord in rat (438, 558, 855), human (215, 565, 595, 1132-1134, 1136), and cat (845). Binding is typically <50% of the strongest labeling observed in the substantia gelatinosa and does not vary between different spinal cord levels. Although there are some inconsistencies between ICC and ISH studies, a general pattern emerges. For AMPA receptor subunits, ICC studies from rat (407, 558, 801, 1206) and human (1367) suggest that GluR4  GluR3/2 > GluR1. Similarly, ISH studies, which distinguish between GluR subunits 2 and 3, suggest that GluR4 = GluR3 > GluR2 > GluR1 (407, 559, 966, 1082, 1145, 1257, 1258), whereas recent studies in human spinal cord suggest low levels of GluR2 subunits (1137) and the complete absence of GluR2 mRNA expression (1366). Low-level GluR2 expression in spinal cord motoneurons is further supported by low GluR2 ICC relative to GluR3 and GluR4 in rat spinal cord (975). Analysis of GluR splice variants is limited to spinal motoneurons where expression of GluR2-flip, GluR3-flip and flop, and GluR4-flip and low levels of GluR1-flop in rat (559, 1238, 1257, 1258) differs from human, where all splice variants are present but flop isoforms predominate (1259).

The distribution of spinal cord kainate receptors has been least studied. KA1 and GluR5 are present in lumbar motoneurons (1257) and GluR5 is present in cervical motoneurons in rat (407). Detection of GluR6/7 and KA2 ICC in cervical motoneurons (976) contrasts with absence of ISH labeling of GluR6, GluR7, and KA2 in lumbar motoneurons (1257) and may reflect differential expression of subunits between cervical and lumbar motoneuron pools, but more likely reflects the greater sensitivity of ICC. There is much speculation on the cellular distribution and function of KA receptor subunits, since, with the exception of cultured hippocampal neurons (728, 729), dorsal root ganglia (528), and cerebellar granule cells (1170), presence of functional high affinity KA receptors in CNS neurons has not been confirmed. Recent developments of specific KA agonists and antagonists should facilitate functional characterization of KA receptors on motoneurons (217, 1201).

NMDAR1 and -2A subunits and transcripts are strongly expressed in cervical and lumbar motoneurons in rat (407, 765, 977, 979, 984,