PLASTICITY OF THE SPINAL NEURAL CIRCUITRY AFTER INJURY

Transcription

1 Annu. Rev. Neurosci : doi: /annurev.neuro Copyright c 2004 by Annual Reviews. All rights reserved First published online as a Review in Advance on February 23, 2004 PLASTICITY OF THE SPINAL NEURAL CIRCUITRY AFTER INJURY V. Reggie Edgerton, 1,2,3 Niranjala J.K. Tillakaratne, 1,3 Allison J. Bigbee, 2 RayD.deLeon, 4 and Roland R. Roy 1,3 1 Physiological Science, 2 Neurobiology, and 3 Brain Research Institute, University of California, Los Angeles, California 90095; 4 Kinesiology and Nutritional Science, California State University, Los Angeles, California 90032; vre@ucla.edu, nirat@lifesci.ucla.edu, abigbee@ucla.edu, rdeleon@calstatela.edu, rrr@ucla.edu Key Words strategies automaticity, motor learning, locomotor rehabilitation, therapeutic Abstract Motor function is severely disrupted following spinal cord injury (SCI). The spinal circuitry, however, exhibits a great degree of automaticity and plasticity after an injury. Automaticity implies that the spinal circuits have some capacity to perform complex motor tasks following the disruption of supraspinal input, and evidence for plasticity suggests that biochemical changes at the cellular level in the spinal cord can be induced in an activity-dependent manner that correlates with sensorimotor recovery. These characteristics should be strongly considered as advantageous in developing therapeutic strategies to assist in the recovery of locomotor function following SCI. Rehabilitative efforts combining locomotor training pharmacological means and/or spinal cord electrical stimulation paradigms will most likely result in more effective methods of recovery than using only one intervention. INTRODUCTION There is a growing awareness that a high level of functional recovery of standing and stepping can be achieved in adult mammals following a complete spinal cord injury (SCI). It is also clear that the level of recovery in motor function is defined, Abbreviations: 5-hydroxytryptamine (5-HT); 2-amino-5-phosphonovaleric acid (AP-5); American Spinal Injury Association (ASIA); brain-derived neurotrophic factor (BDNF); central nervous system (CNS); central pattern generation (CPG); electrical stimulation (ES); electromyography (EMG); extensor digitorum longus (EDL); gamma-aminobutyric acid (GABA); glial fibrillary acidic protein (GFAP); glutamic acid decarboxylase (GAD); long-term potentiation (LTP); long-term depression (LTD); m-chlorophenylpiperazine (m- CPP); medial gastrocnemius (MG); neurotrophin-3 (NT-3); noradrenaline (NA); N-methyl- D-aspartate (NMDA); paw contact (PC); peripheral nervous system (PNS); soleus (SOL); spinal cord injury (SCI); swing-phase force field (SWPFF); tibialis anterior (TA) X/04/ $

2 146 EDGERTON ET AL. to a significant degree, by the level and types of motor training or experience following the injury (Edgerton et al. 2001a, Wernig et al. 1995). These findings suggest strongly that any promising interventional strategies for regeneration and redevelopment of supraspinal-spinal connectivity following SCI are likely to be more successful when the intervention is combined with some type of motor training. Insight into the mechanisms of recovery depends on a basic understanding of the neural control of standing and stepping in the uninjured compared to the injured spinal cord, as well as the properties and adaptations of the force generating tissue, i.e., muscle-tendon units. Much of the insight we have gained in the neural control of locomotion is attributable to the fact that the basic networks that control locomotion in animals that walk, fly, and/or swim are remarkably similar. The physiological properties of the supraspinal and spinal circuits that control posture and locomotion are similar across many vertebrate systems. Glutamate and glycine are essential neurotransmitter systems for central pattern generation (CPG) in virtually all vertebrates. Similarly, noradrenergic and serotonergic neurotransmitter systems are important modulators of the basic rhythmic cycle associated with locomotion among most vertebrates (Grillner 2003). The extensive conservation of the neuromotor control system and its neuromodulators and neurotransmitters enable us to gain a deeper understanding of the functional properties of complex systems by studying simple animal models (Grillner 2003). This in turn has also been an important factor in helping to understand the recovery of locomotion following SCI. In this review we highlight some features of the normal neuromotor control apparatus and how SCI changes this neural control. An understanding of this control provides a reasonable rationale for the level of plasticity that persists and the highly functional state of posture and locomotion that can be achieved in the neuromusculoskeletal system following SCI. Because the number of papers published on the topic of SCI over the past few years has risen rapidly, we restricted the present review primarily to those papers that include a complete SCI and motor training post-sci. The two objectives of this review are (a) to summarize the level of motor or functional control that can persist in the lumbosacral spinal cord in the absence of supraspinal control, and (b) to provide some insight into the physiological and biochemical changes that can occur in the lumbosacral segments, and the sensory and motor adaptations that can account for locomotor recovery after SCI. A number of recent reviews have been written on similar topics (Barbeau et al. 2002; Edgerton et al. 2001a; Edgerton & Roy 2002; Grillner 1981, 2003; Grillner & Wallen 1985; Harkema 2001; Rossignol et al. 2002a,b; Wolpaw & Tennissen 2001). AUTOMATICITY IN POSTURE AND LOCOMOTION: SOME BASIC NEUROBIOLOGICAL PRINCIPLES OF MOTOR CONTROL BEFORE AND AFTER SCI In the uninjured state, the spinal cord, along with the brain, plays a considerable role in the automaticity of motor control (Baev 1998, Edgerton et al. 2004, Orlovsky & Feldman 1972). Automaticity describes the concept by which one is

3 PLASTICITY AFTER SCI 147 able to carry out complex but routine motor tasks without conscious thought, such as walking across a room. Descending brain control, central pattern generator activity, and peripheral inputs are components of automaticity. Following SCI, the spinal circuitry below the lesion site does not become silent; rather it continues to maintain active and functional neuronal properties. The potential for the spinal circuitry to generate oscillating coordinated motor patterns remains, and peripheral input continues to flow into the spinal cord and be processed after a complete SCI, although in a modified manner. Spinal automaticity is then comprised of CPG circuitry combined with sensory input from the periphery. One can ask, how did automaticity with such utility evolve? All animals on Earth with vastly different neuromusculoskeletal structures have evolved within a 1 g gravitational environment. Over millions of years the execution of many complex postural and locomotor tasks has evolved to become automatic in many ways within this constant environment (Edgerton et al. 2001b). The automaticity of the neural control of locomotion and posture facilitates the execution of complicated tasks that might otherwise require decision-making processes that would take additional time. Automaticity also frees selected neural networks that can be reserved for less predictable events. From this perspective spinal automaticity seems to have been learned in the evolutionary process, i.e., evolutionary learning has played a key role in shaping the mammalian neural systems that control posture and locomotion. A few obvious examples of the automaticity that is built into the neural control of posture and locomotion are shorter steps taken by the inside limb when walking in an arc or by the limb of the side of the body to which the head is turned; increasing the level of excitation of extensor muscles when the body is bearing more load; and enhancing flexion of the ipsilateral leg when stepping over an object versus enhancing extension in the contralateral limb. In the extreme case of SCI where the spinal cord is completely severed in the midthoracic region, there is no supraspinal control. In this case the automaticity within the spinal cord becomes even more critical for the CPG circuitry to successfully process sensory input. The manner in which these features of the control system are affected after a complete SCI is discussed below. Supraspinal Control of Posture and Locomotion Shik & Orlovsky (1976) proposed a two-level automatism control system for locomotion. One level provides nonspecific tonic input, which determines the speed of locomotion. The second level is responsible for making fine adjustments in the control of the limbs, including maintaining equilibrium, by interacting with multiple modes of sensory information such as proprioceptive, visual, and auditory inputs (Figure 1). Normally there are several descending spinal tracts that project motor commands from the brain to the spinal cord, including the reticulospinal, vestibulospinal, rubrospinal, and corticospinal tracts. Each of these tracts receives multiple modes of input and incorporates peripheral sensory modalities with information from the cerebellum and other supraspinal structures to produce appropriate locomotor responses in many different situations.

4 148 EDGERTON ET AL. The specific control features of each of the descending spinal tracts in controlling locomotion are poorly understood. Also the anatomy of these tracts differs to some degree from species to species. Although the detailed clinical manifestation of a SCI will reflect the amount of tissue damage, it seems that the specific tracts that are damaged are also an important factor. However, there has been relatively little focus on how the loss of specific tracts following SCI relates to the specific kinds and levels of motor function that remain after the injury. For example, the corticospinal tract of quadrupedal animals does not play an essential role in generating the basic locomotor pattern, but it is very important in making fine adjustments (Eidelberg 1981, Grillner 2003). Although damage to this tract will not result in a severe deficit during treadmill locomotion, these fine adjustments in the basic activation patterns during locomotion in a variable environment can be important. In nonhuman primates, however, the basic locomotor patterns can be relatively normal without corticospinal input, with most of the dysfunction occurring primarily in the distal musculature, i.e., in the wrist, ankle, and digits (Eidelberg 1981, Hodgson et al. 2003). After a more severe SCI, however, the integral role of spinally mediated mechanisms in postural and locomotor control becomes more apparent (Figure 1B). Spinal Control of Posture and Locomotion CENTRAL PATTERN GENERATION CPG is a physiological phenomenon in which oscillatory motor output is generated in the absence of any oscillatory input. In mammalian, nonhuman systems, CPGs within the lumbosacral spinal cord segments represent an important component of the total circuitry that generates and controls posture and locomotion. With an appropriate pharmacological stimulus, a highly coordinated oscillatory pattern, i.e., alternating flexion and extension (fictive locomotion), can be generated in a fictive preparation by the circuitry within the spinal cord in the absence of input from the brain or from the periphery. During this oscillatory behavior, interneurons are active throughout all layers of the gray matter of the spinal cord and over all of the lumbosacral segments (Edgerton et al. 1976, Jordan & Schmidt 2002). Many excellent reviews on CPG and locomotion have been written (Barbeau et al. 2002; Dietz 2003; Edgerton et al. 2001a; Grillner 2002, 2003; Grillner & Wallen 1985; Harkema 2001; Rossignol et al. 2002a,b; Wolpaw & Tennissen 2001) and provide more detailed information. Although there are reports that claim to demonstrate CPG in nonhuman primates (Fedirchuk et al. 1998) and in humans (Calancie et al. 1994, Dimetrievic et al. 1998, Gurfinkel et al. 1997,Gerasimenko et al. 2002), the evidence is not conclusive, largely because of the infeasibility of eliminating all peripheral as well as supraspinal input to the spinal cord, at least in humans. SENSORY INPUT An intact and functional neuromotor control system cannot rely solely on CPG or on direct supraspinal control to generate effective movements, particularly in a dynamic environment. The profuse sources of peripheral sensory

5 PLASTICITY AFTER SCI 149 information represent an integral component of the neuromotor control system, allowing the spinal cord to effectively interface with its physical environment. The spinal cord receives sensory input from all of the mechanoreceptors and cutaneous receptors and produces the appropriate temporal response, thus generating many highly adaptable motor tasks. In this context, it is convenient to think of the spinal cord as interpreting the ensemble of sensory information at any given time, as opposed to responding to each receptor in a stereotypically reflexive manner. This feature of the spinal cord makes it smart and provides the foundation for a substantial portion of the recovery of postural and locomotor function following SCI. The intrinsic CPG activity and the sensory input to the spinal circuitry combined allow for rapid adjustment to parameters such as the speed of stepping, the level of load imposed during stepping, and a wide range of unpredictable sensory anomalies (Edgerton et al. 2001a). We propose that the spinal cord processes and interprets proprioception in a manner similar to how our visual system processes information. When we view a painting, the brain interprets the total visual field, as opposed to processing each individual pixel of information independently, and then derives an image. At any instant the spinal cord receives an ensemble of information from all receptors throughout the body that signals a proprioceptive image that represents time and space, and it computes online which neurons to excite next based on the most recently perceived images. Those CPG neurons that generate locomotor patterns appear to predict the next sequence of neurons to activate on the basis of the specific groups of neurons that were just activated. The importance of the CPG is not simply its ability to generate repetitive cycles, but also to receive, interpret, and predict the appropriate sequences of actions during any part of the step cycle, i.e., state dependence. The peripheral input then provides important information from which the probabilities of a given set of neurons being active at any given time can be finely tuned to a given situation during a specific phase of a step cycle. An excellent example of this is when a mechanical stimulus is applied to the dorsum of the paw of a cat. When the stimulus is applied during the swing phase, the flexor muscles of that limb are excited, and the result is enhanced flexion. However, when the same stimulus is applied during stance, the extensors are excited (Forssberg et al. 1975). Thus, the functional connectivity between mechanoreceptors and specific interneuronal populations within the spinal cord varies according to the physiological state. Even the efficacy of the monosynaptic input from muscle spindles to the motoneuron changes readily from one part of the step cycle to another, according to whether a subject is running or walking (Simonsen et al. 1999). The injured spinal cord interprets sensory changes in load and speed The spinal cord clearly interprets ensembles of information derived from mechanoreceptors in the musculoskeletal system and activates the appropriate motor pools in a precise and highly coordinated manner in response to changes in speed and load. An example of the spinal cord s ability to receive complex proprioceptive information and use it in a functional way is shown in Figure 2.

6 150 EDGERTON ET AL. Figure 2 Loading alters the activation levels of motor pools during stepping in human SCI patients. The relationships between soleus EMG mean amplitude (µv) and limb peak load (N) from a nondisabled subject (A, ND-1) and a subject with a SCI classified on the ASIA scale as an A, i.e., having a complete injury (B, SCI-A1), over a range of loading conditions are shown. Each data point represents one step, and each symbol represents a series of consecutive steps at one level of body-weight support. (Reproduced with permission from Edgerton et al. 2001a.) The activation level of an extensor muscle, in this case the soleus, is modulated according to the amount of load placed on the lower limbs of a nondisabled human subject (Harkema et al. 1997). The increase in activation (Figure 2A), illustrated by the amplitude of the electromyographic (EMG) signal, is directly related to the load imposed on the limb. Data from a similar experiment in a complete SCI patient is also shown (Figure 2B). Even though the SCI subject has no voluntary control of any muscles and no sense of sensation from tissues below the lesion, there is a clear similarity in the relationship between the level of loading and the level of motor pool activation as reflected by the EMG amplitude compared to the nondisabled subject. Similar adaptations have been observed with respect to speed (Forssberg et al. 1980a,b) and force modulation (Figure 3; Lovely et al. 1990) in low thoracic, chronically spinalized cats during assisted treadmill locomotion. Indirect evidence suggests that the spinal cord neural circuits vary the speed of locomotion primarily by modulating the level of excitation of the extensor muscles. For example, there are greater angular excursions, enhanced muscle forces, and shorter EMG durations in the SOL and MG at a faster (Figure 4B) than a slower (Figure 4A) treadmill speed in complete SCI cats (Figure 4; Lovely et al. 1990). However, experiments in decerebrate cats showed that if treadmill speed is kept constant, increased input into the midbrain has no effect on the duration of the stance phase or step frequency (Shik et al. 1966). This latter observation suggests that the cycle duration is influenced by mechanical factors such as the position or the placement of the hindlimb and, therefore, is mediated by proprioceptive signals.

7 PLASTICITY AFTER SCI 151 Figure 3 Hindlimb muscle EMG and force production (via tendon force transducers) are prominent during assisted treadmill locomotion in adult spinal cats. Comparisons of the activation patterns of the soleus (SOL) and medial gastrocnemius (MG), and corresponding force production in a normal (A) and a complete low thoracic spinal cat that has been trained to step on a treadmill with full weight-bearing (B) are shown. The beginning of the stance phase (paw contact; PC) and the beginning of the swing phase (ankle flexion; Fa) are noted. The EMG bursts per step are less robust, and there is a slightly more irregular and more brief force pattern in both the SOL and MG after than before SCI. The peak force levels are also lower after than before SCI, particularly in the MG, which has a higher percentage of larger motoneurons that require more excitatory drive to be activated. However, there are significant forces and activation levels in the muscles of this complete spinal cat. The bold bars indicate the period of contralateral support. (Reproduced with permission from Lovely et al )

8 152 EDGERTON ET AL. Figure 4 EMG and force production are altered by changing the speed of assisted treadmill locomotion in complete SCI cats. EMG and force patterns of the soleus (SOL), medial gastrocnemius (MG), tibialis anterior (TA), and extensor digitorum longus (EDL) of an adult cat with a complete SCI during stepping at a slow speed (A; 0.1 m sec 1 ) and at a faster, but modest, speed (B; 0.6 m sec 1 ) on a motor-driven treadmill. Clonus is more evident in the EMG and force records of these muscles at the slower than at the faster speed of stepping. Also note the shorter stance period at the faster speed, indicating the modulation of extensor power output (greater recruitment for a shorter period) to walk faster. The vertical lines at the beginning of the SOL force indicate the beginning of stance, i.e., paw contact (PC), as derived from high-speed film (200 fps). Note the different scales for the force and EMG records in A and B. Hip flexion, Fh; knee flexion, Fk; ankle flexion, Fa; knee extension prior to PC, E1k; ankle extension prior to PC, E1a; hip extension, Eh. (Reproduced with permission from Lovely et al )

9 PLASTICITY AFTER SCI 153 SCI patients can voluntarily initiate locomotion An understanding of the extensive level of built-in spinal automaticity can be used in developing therapeutic strategies to facilitate functional recovery after SCI. With the increasing success in improving the locomotor potential of laboratory animals, however, the question has been raised as to the functional significance of this motor recovery in humans after a complete SCI, since there is presumably no voluntary initiation of stepping or standing. This presumption, however, is not entirely correct. Both standing and stepping can be initiated voluntarily, although indirectly, in SCI patients. For example, when an ASIA A SCI patient is standing with bilateral weight support, stepping can be initiated (consciously) by shifting the body weight to one leg and moving the head and trunk so that the hip position of the contralateral leg is extended by leaning the body forward (Wernig & Müller 1992). Proprioceptively, this maneuver is similar to unloading a limb at the end of the stance phase of a step, which facilitates the initiation of the swing phase of that leg (Duysens & Pearson 1980, Harkema 2001, Wernig & Mueller 1992). The same events initiate stepping on a moving treadmill belt because the weight-bearing limb moves backward into hip extension. A moving treadmill, however, is not required to successfully initiate stepping because the patient can use the upper body to alternate loading from one side to the other at the appropriate time to facilitate the joint actions bilaterally. THE INJURED SPINAL CORD IS AN ALTERED SPINAL CORD After a SCI, supraspinal and spinal sources of control of movement differ substantially from that which existed prior to the injury, thus resulting in an altered spinal cord. The concept of an altered spinal cord after SCI implies that the spinal cord processes input and generates motor output in a different manner as a result of injury-related adaptations (Edgerton et al. 1997b, 2001b). How, then, does the automaticity of posture and locomotion emerge from the interactions between the sensory inputs and the CPG circuitry in spinalized animals? For these two systems to work in synergy, each system must have intrinsic activation and inhibition patterns that can generate coordinated motor outputs. For example, the sequence of muscle activation patterns associated with stepping can occur only if a critical level of synergy occurs between the peripheral inputs and the cyclic events of the CPG. At any given time bin within a step cycle, the sensory information interfaces with the CPG circuitry to (a) apprise the CPG circuitry about the present state of the step cycle, (b) inform the CPG circuitry about which bin(s) preceded that instant, and (c) predict which bin(s) should occur next. In this scenario a bin consists of that sensory input and the stage of activation-inhibition of the neurons that generate the CPG output at that instant. The temporal patterns of peripheral inputs must be matched with the intrinsic CPG activity for locomotion to continue effectively. After a complete SCI, the interaction between the CPG circuitry and the peripheral inputs is critical because a major source of control, the brain, has been eliminated.

10 154 EDGERTON ET AL. Spasticity is a sign of activity in the spinal circuitry Spasticity might be considered as automaticity gone awry. Although extensive muscle contractions occur with spasticity, there is little reciprocal activation of agonistic and antagonistic motor pools, as occurs during locomotion (de Leon et al. 1998a). One possible explanation for a more-or-less random motor pool activation is that the spinal cord has lost the ability to synchronize and interpret the coordinated ensemble of afferent information that produces a predictable motor outcome in nondisabled subjects. This deficit may be due to the absence or rare occurrence of synchronized events normally associated with load-bearing stepping. In the absence of these coordinating events, the spinal cord loses the ability to synchronize input into functional movements of the limbs. The presence of spasticity is, however, a positive sign of the potential for a SCI patient to regain some locomotor ability. A patient who is hyporesponsive to sensory input is less likely to respond to the proprioceptive input associated with load-bearing stepping. Clearly, understanding the physiological mechanisms that underlie spasticity will enhance our efforts to facilitate locomotor recovery in SCI patients. MOTOR OUTPUT IS ENHANCED BY REPETITIVE TRAINING Chronic Motor Training Modulates Spinal Plasticity to Enhance Motor Output After SCI Experiments with spinal cats using chronic locomotor training paradigms suggest that the ability to learn and successfully perform a motor task is dependent on repetitive practice (de Leon et al. 1998a,b; Lovely et al. 1986) and appears to be specific to the task being tested (de Leon et al. 1998a,b). Although training for a specific motor task such as stepping can enhance the capacity to perform that task, it may reduce the capacity to perform a different complex motor task that uses different patterns of information, e.g., standing (de Leon et al. 1998b). Cats that were trained to step perform that motor task well, whereas those trained for weight-bearing standing did not step well. These findings imply that the functional state of the cord is shaped by specific locomotor training regimens. Although the mechanisms underlying locomotor training-enhanced plasticity are not well understood, it is clear that the physiological state of the cord can be affected by activity-dependent processes that can influence its ability to learn and perform a specified motor behavior. The Spinal Cord Can Respond to Novel, Acute Perturbations Although spinal cord plasticity can be modulated to alter its motor capacity after a complete SCI, one may argue that this behavior requires long periods of motor training and new neuronal connections to form within the spinal cord. However, the state-dependent property of the spinal circuitry, which enables highly functional

11 PLASTICITY AFTER SCI 155 Figure 5 Swing trajectories of the foot (lower traces) and ankle (upper traces) ofa spinal cat responding to the placement of a 1-cm diameter bar in the path of the foot during the swing phase of the step cycle. The control trace shows the trajectory before placement of the bar, the trip trace shows the trajectory when the cat hit the bar, and the remaining trace shows the trajectory when the foot cleared the bar. The bar position for each frame of the digitized video is shown by an open circle. Note the increased elevation of the foot well before it encountered the bar in the step where the rod was avoided. (Reproduced with permission from Hodgson et al ) and integrated responses to occur, can also occur in novel, acute situations. For example, when an object is placed in front of a spinal cat stepping on a treadmill during the swing phase of one hind limb, that limb will exhibit a greater degree of flexion during the following steps to avoid the perturbation (Forssberg et al. 1975). This hyperflexion persists for several steps even after the removal of the perturbation (Figure 5; Hodgson et al. 1994, Nakada et al. 1994), which suggests that a learning and memory-type phenomenon may be taking place. This must be considered as more than a momentary adaptation because a memory trace is shown behaviorally in a number of steps immediately following the removal of the perturbation. More recent evidence of the smartness of the spinal cord was demonstrated when a robotic device was used to apply specific forces to rat hindlimbs during certain phases of the step cycle. In one experiment a downward force proportional to the velocity of stepping was applied unilaterally or bilaterally to the ankle(s) of spinal rats. Step timing and kinematics were altered within a few steps, enabling locomotion to continue (Timoszyk et al. 2002). In a separate but similar set of experiments, a robotically induced upward force proportional to the forward velocity of the swing (SWPFF) was exerted at the ankle of one limb during the swing phase, resulting in a visually obvious kinematic disturbance. After as few as 20 steps, the limb adjusted its output to become more kinematically

12 156 EDGERTON ET AL. Figure 6 The spinal cord adapts in a functionally appropriate manner to an acute perturbation to normalize step kinematics. A robotic system applied an upward force (swing-phase force field; SWPFF) at one ankle of a complete spinal rat during bipedal, weight-supported stepping on a treadmill. The rat took 20 steps with the SWPFF on and 20 with the SWPFF off for 4 consecutive bouts (between total steps). The first and fourth SWPFF-on and -off bouts are shown. (A) OFF-1 swing phase duration is highly consistent (mean ± SD = 551 ± 127 ms). (B) OFF-4 is similar to OFF-1 (683 ± 331 ms). (C) ON-1 SWPFF exposure increases the swing-phase duration (1103 ± 610 ms). (D) ON-4 swing duration shifted toward OFF-1 and OFF-4 levels (582 ± 326 ms). Thus, step characteristics return toward normal within a short period of time (about 4 min total) upon repeated exposure to the SWPFF perturbation. similar to that observed prior to the perturbation, although with different rectus femoris EMG patterns (de Leon et al. 2002). Furthermore, with repetitive trials of 20 steps on-, 20 steps off 4 bouts of the SWPFF paradigm, the average swing duration during the force-field-on period decreased to force-field-off levels by the fourth bout (Figure 6; Liu et al. 2003). These studies show that, in essence, the spinal cord is solving problems in real time based on the continually changing state of incoming peripheral information to elicit a nearly constant behavior, even though the means to the endpoint differ. These studies imply that spinal learning can occur in a very short period of time, and a type of memory trace allows for quicker adaptation upon reexposure to a given perturbation. Although the underlying cellular mechanisms are unknown, studies are in progress to determine if some of the molecules and processes involved in spinal learning are similar to those involved in hippocampal learning, and whether this phenomenon continues for longer periods (hours to days). Indeed, long-term potentiation (LTP) and long-term depression (LTD), which are vital in hippocampal learning, are also exhibited in dorsal horn neurons in the spinal cord in response to nociceptive stimuli (Garraway & Hochman 2001, Ikeda et al. 2003, Rygh et al. 2002). These observations confirm that hippocampal learning like phenomena

13 PLASTICITY AFTER SCI 157 can occur in the spinal cord. Whether spinal motor learning occurs via similar processes or affects spinally mediated pain responses remains to be determined, but it will be an interesting topic for future study. Biochemical and Pharmacological Evidence for Spinal Cord Plasticity After Injury As shown above, activity-dependent motor training facilitates the recovery of posture and locomotion after a complete SCI in mammals. Because the functionally recovered spinal animals showed no evidence of regeneration of descending pathways (Joynes et al. 1999) or showed minimal changes in hindlimb skeletal muscle properties (Roy et al. 1999, 1998; Roy & Acosta 1986) to account for the recovery characteristics, the functional behavior exhibited by these animals must have been mediated by the plasticity in existing spinal pathways. This plasticity may occur at any of many spinal cord regions or cell types such as motoneurons, premotor pattern-generating neurons, and/or nonneuronal cell types. There could also be anatomically altered synaptic connections, increased active zones of synapses, altered sensitivities of neurotransmitter receptors, or altered production of neurotransmitters. Pharmacologically induced activation of the spinal cord has been well studied. In the isolated neonatal rodent spinal cord, fictive locomotion can be activated by application of different neurotransmitters such as N-methyl-D-aspartate (NMDA), 5-hydroxytryptamine (5-HT), or dopamine, resulting in a stable rhythmic motor output (Cazalets et al. 1998, Kiehn & Kjaerulff 1998, Kudo & Nishimaru 1998, Schmidt et al. 1998). These patterns closely resemble the activation of hindlimb muscles during locomotion in the intact adult rat (Roy et al. 1991). These data showed a significant role for these neurotransmitter systems in facilitating locomotor activity and thus suggest that these agents might be useful for inducing locomotion in SCI animals. Indeed, the alpha-2-noradrenergic (NA) agonist clonidine induced stepping early (within 1 week postinjury) in adult, nontrained spinal cats that could not otherwise step (Chau et al. 1998). The glutamatergic agonist NMDA had a dramatic positive effect on locomotion at a later stage (months postinjury) in chronic spinal cats. This effect was blocked by the NMDA antagonist Ap5 (Giroux et al. 2003), which demonstrated that the activation of NMDA receptors may play a role in facilitating stepping. Ap5 was, however, ineffective in the induction of stepping in acute spinal cats (Chau et al. 2002), which suggests that the cellular biochemistry of the spinal cord changes over the time course of an injury and is an important consideration when developing therapeutic strategies for SCI. Administration of 5-HT, its precursor 5-hydroxytryptophan (5-HTP), or the 5-HT agonists, quipazine and m-chlorophenylpiperazine (m-cpp), are also effective in improving locomotion in spinal cats (Barbeau & Rossignol 1990, 1991) and rats (Feraboli-Lohnherr et al. 1999, Kim et al. 2001), even though descending 5-HT axons are eliminated by a complete SCI. Consistent with these pharmacological studies, autoradiographic receptor-binding studies showed elevated levels

14 158 EDGERTON ET AL. of alpha1- and alpha2-na and 5-HT1A receptors in selected laminae of the lumbar spinal cord segments of spinal cats at days post-sci, which gradually returned to control values (Giroux et al. 1999, Roudet et al. 1996). It is not known whether the locomotor patterns are stimulated by directly activating the neurons responsible for CPG or by modulating the spinal circuits that process proprioceptive information. A complete SCI, and step or stand training after SCI, can markedly change the physiological and biochemical state of multiple neurotransmitter-modulator systems, and thereby change the pharmacological properties of the sensorimotor pathways that generate stepping and standing. Thus, the response of spinal animals to externally administered drugs that mimic or antagonize various neurotransmitter molecules provides a clue for the changes in the synaptic milieu (de Leon et al. 1999b). Pharmacological as well as biochemical data show that task-specific adaptive changes occur in spinal neuronal circuits of spinal cats trained to stand or step (de Leon et al. 1999b; Edgerton et al. 1997a; Rossignol et al. 2001; Tillakaratne et al. 2000, 2002). Administration of modest doses of strychnine, a glycine receptor antagonist that did not affect locomotion in step-trained spinal cats, significantly improved the stepping ability of poorly stepping, nontrained or stand-trained spinal cats (de Leon et al. 1999b). Similarly, the stepping ability of poorly stepping spinal cats can be dramatically improved by administration of the GABA A receptor antagonist bicuculline (Edgerton et al. 1997a). The improvement of stepping upon the administration of strychnine or bicuculline may occur by facilitating neuronal excitation by blocking the abnormally high levels of general inhibition resulting from a complete SCI (de Leon et al. 1999b). Nontrained spinal rats have increased GABA synthetic enzyme GAD 67 and glycine and GABA A receptors in the lumbar spinal cord, whereas step-trained spinal rats have near-normal levels (Edgerton et al. 2001a, Tillakaratne et al. 2000). In agreement with this, GAD 67 protein levels in lumbar spinal cord extracts were higher in spinal cats who were step trained for two months and then not trained for five months (de Leon et al. 1999a, Tillakaratne et al. 2002). After just one week of retraining, however, there was a significant reduction in GAD 67 protein levels (Tillakaratne et al. 2002). In step-trained cats, the GAD 67 levels around motoneurons were inversely correlated with stepping ability (Tillakaratne et al. 2002). Training spinal cats to weight-bear also appears to reduce GABA signaling in some spinal interneurons. Cellular mapping of GAD 67 also suggests that the spinal cord of both stand-trained and nontrained spinal cats have selectively higher inhibitory potential compared to step-trained animals. For example, nontrained and stand-trained cats had more inhibitory GABAergic inputs (increased number of GAD 67 -positive terminals) around the motoneurons than the step-trained cats. Furthermore, spinal cats trained to stand on one leg had elevated GAD 67 within the motor pools corresponding to knee flexors in the trained leg but not in the nontrained leg. These findings suggest that the inability of stand-trained spinal animals to step is closely linked to an elevated level of inhibition of flexor motor pools. The specificity of GABAergic changes associated with the specific type of training emphasizes the concept that the activity in the neural networks of

15 PLASTICITY AFTER SCI 159 the lumbar spinal cord is, to a large extent, determined by the pattern of activity that the hindlimbs experience. How these changes are related to increased inhibitory and/or excitatory synaptic activity, and in which neural circuits this is manifested, remains unclear. SCI also may have an effect on nonneuronal cell types in the spinal cord. For example, compared to control rats, spinal rats showed elevated immunostaining of the GABA A receptor gamma-2 subunit in astroglial cells, i.e., cells positive for the cellular marker glial fibrillary acidic protein (GFAP) in the ventral horn regions surrounding retrogradely labeled SOL and TA motoneurons (Bravo et al. 2002). This elevated level returns toward nearly undetectable control levels after six weeks of step training. Plasticity of spinal circuits may also be mediated by activity-dependent induction of neurotrophins. For example, voluntary exercise induces an upregulation of BDNF and NT-3 mrna and protein levels in the spinal cord (Gomez-Pinilla et al. 2002, 2001; Ying et al. 2003). Neurotrophins stimulate axonal growth through the lesion site of an injured spinal cord (Tuszynski et al. 1994), and exogenously applied BDNF can improve stepping in spinal rats (Jakeman et al. 1998). Whether BDNF acts directly on those neurons that induce CPG or whether it plays a facilitating role, such as strychnine and bicuculline, is not known. Effects of Electrical Stimulation on Locomotor Recovery After SCI Numerous experiments have demonstrated that nonpatterned electrical stimulation (ES) of the lumbosacral enlargement can induce locomotor patterns and even hindlimb stepping in acute and chronic low-spinal kittens and acute spinalized rats (Iwahara et al. 1992), cats (Edgerton et al. 1976, Grillner & Zangger 1984), and humans (Dimitrijevic et al. 1998; Gerasimenko et al. 2003, 2002). ES differs from direct stimulation of individual muscles via electrodes implanted intramuscularly or around/near peripheral nerves. Only the spinal cord ES approach capitalizes on the coordinating capacity that is built into the spinal circuitry, which remains largely functional after most SCI. Dimitrijevic et al. (1998) showed that bilateral or unilateral epidural spinal cord ES can elicit step-like EMG activity bilaterally or unilaterally, respectively, in chronic, complete paraplegic subjects (ASIA A). Increased stimulation amplitude resulted in increased EMG amplitudes and an increased frequency of rhythmic activity. High frequencies of stimulation (>70 Hz) produced tonic activity in the leg musculature, which suggests that the upper lumbar stimulation may activate neuronal structures that then recruit interneurons involved in CPG. Gerasimenko et al. (2002) also used spinal cord stimulation in complete SCI patients (thoracic or cervical level) and found that stimulation of the ventral surface of the cord initially produced tonic activity in the motor pools of the leg, eventually giving way to rhythmic activity. The most effective electrode placement for rhythmic activity was at L2, and once begun, it continued throughout the stimulation

16 160 EDGERTON ET AL. period and persisted for 1 or 2 cycles after cessation of the stimulus. Kinematically, the movements resembled bicycling, but the frequency and coordination pattern corresponded to stepping. Interestingly, progressive increases in the stimulation intensity produced activation from the proximal to the distal muscles and from tonic to rhythmic activity. The stimulation site and parameters for generating tonic and locomotor-like activity with spinal cord ES in chronic spinal (T10) cats were generally similar to that in humans (Gerasimenko et al. 2003, 2002). Dorsolateral column lesion experiments and unilateral dorsal rhizotomy experiments resulted in loss of locomotor activity and rhythmic activity of the unlesioned side, respectively, when stimulating, which indicates that afferent inputs, i.e., dorsal root afferents and propriospinal fibers, can play a significant role in initiating and maintaining locomotor activity (Dimitrijevic et al. 1998, Gerasimenko et al. 2002). Dimitrijevic and colleagues suggest that stimulation of the same lumbar site at different frequencies has distinctly different physiological effects (Pinter et al. 2000). For example, stimulation at Hz was effective in controlling spasticity. Stimulating at Hz at 7 10 V initiated and maintained rhythmical step-like hip and knee flexion in an ASIA A subject (Dimitrijevic et al. 1998). In addition only extensor movements were produced when stimulating at <15 Hz. Thus, the evidence to date seems to suggest that nonspecific stimulation of the upper lumbar segments with epidural electrodes, combined with the sensory input generated by weight-bearing stepping, has considerable potential as a therapeutic strategy to improve locomotor performance in SCI patients (Barbeau et al. 1999). Herman et al. (2002) studied the efficacy of spinal cord ES in facilitating functional gait in an ASIA C patient. They found that motor training alone improved the stepping pattern and reduced spasticity, but overground stepping was very slow and tiring. A combination of treadmill training and stimulation, however, markedly improved the quality and quantity of stepping during the training session and resulted in immediate improvement in the quality of overground walking. These encouraging results imply that a generalized, nonspecific electrical stimulation of upper lumbar neurons may be sufficient to generate, or at least facilitate, rhythmic stepping movements of the lower limbs. The results raise the possibility that when spinal cord ES is combined with proprioceptive input associated with weightbearing stepping, greater levels of activation and coordination can be attained than if either method is applied alone. Electrical microstimulation of the spinal cord has been used to selectively activate specific hindlimb motor pools to produce predictable multi-joint movements. Multiple fine stimulating electrodes are inserted intraspinally in activation pools located in the ventral horn and activate specific muscle groups independently. The feasibility of using this approach to produce functional motor tasks was studied in chronically implanted (lumbar enlargement, 6 months) intact cats (Mushahwar et al. 2000). Stimulation through selective electrodes was used to produce single and multi-joint movements, in some cases producing a standing posture in the intact cat. In spinally intact anesthetized cats, bilateral locomotor-like activity was

17 PLASTICITY AFTER SCI 161 elicited via stimulation of a few electrodes placed in the ventral horn of the spinal cord, particularly in lamina IX. Whether intraspinal microstimulation of specific activation pools involves the CPG circuitry is unknown at this time. The current strategies to electrically stimulate the spinal cord to initiate, modulate, and control locomotion, show that (a) locomotor-like movements of the limbs can be generated after a complete SCI via electrodes placed over the dura of the upper/middle lumbar segments; (b) microelectrodes placed in localized regions within the spinal cord gray matter can elicit more defined movements of the entire limb; and (c) spinal cord ES may prove to be more efficacious when combined with other interventions such as step training and pharmacological facilitation. HUMAN SCI: A PERSPECTIVE Three general principles have emerged largely from the study of spinal cats: (a) body weight supported treadmill training improves the ability of the lumbosacral spinal cord to generate weight-bearing stepping; (b) patterns of sensory input provided during locomotor training are critical for driving the plasticity that mediates locomotor recovery; and (c) pharmacological treatments can be used to excite the spinal neurons that generate stepping. The question of whether the human spinal cord is capable of the level of adaptability demonstrated in other mammalian systems has not been fully answered. However, there is increasing evidence that the human spinal cord is capable of a significant amount of plasticity and that this plasticity is, to a large extent, driven by activity-dependent processes. A number of clinical studies demonstrate the applications of these lessons learned from the study of spinal cats toward facilitating the recovery of walking in humans with a complete SCI (Curt et al. 2004). Treadmill Training and the Recovery of Walking Ability After SCI In humans with incomplete SCI, treadmill training has a profound impact on overground walking ability (Wernig et al. 1995). To date, however, there has not been a similar training effect on overground walking in humans with clinically complete SCI. There is evidence, however, that treadmill training can improve several aspects of walking on a treadmill with some weight-supporting assistance in humans with clinically complete SCI. Dietz and colleagues (1995) reported that after several weeks of treadmill training, the levels of weight bearing that can be imposed on the legs of clinically complete SCI subjects during treadmill walking significantly increases. When stepping on a treadmill with body-weight support, rhythmic leg muscle activation patterns can be elicited in clinically complete subjects who are otherwise unable to voluntarily produce muscle activity in their legs (Maegele et al. 2002). A recent study has demonstrated that the levels of leg extensor muscle activity recorded in clinically complete SCI subjects significantly improved over

18 162 EDGERTON ET AL. the course of several weeks of step training (Wirz et al. 2001). Interestingly, the levels of extensor muscle activity decreased over a three-year period of time following the training program. Together, these results indicate that a use-dependent phenomenon may exist for the human spinal cord as has been reported for spinal cats (de Leon et al. 1999a,b). In short, the stepping ability of clinically complete SCI subjects can improve in response to step training, but the level of improvement has not reached a level that allows complete independence from assistance during full weight-bearing. Use of Pharmacological Therapies in Enhancing Walking after SCI Some of the pharmacological agents used successfully to induce weight-bearing stepping in spinal cats have also been used in humans with SCI in the hopes of enhancing walking performance, but have produced somewhat disappointing results. The NA agonist clonidine, for example, which is highly effective in inducing full weight-bearing stepping in spinal cats (Barbeau & Rossignol 1987, Chau et al. 1998) has had minimal effects on treadmill walking performance (Barbeau & Norman 2003) and can even have a depressing effect on EMG activity levels during treadmill walking (Dietz et al. 1998a,b) in clinically complete SCI patients. Although any number of factors may account for the lack of an effect of clonidine in humans with complete SCI, clonidine may have a beneficial effect in humans if it is combined with treadmill training, and its effects may depend on the time after the injury. In support of this, Chau and colleagues (1998) reported that a combined treatment of clonidine and treadmill training in spinal cats beginning two days after spinalization had a significant facilitatory effect on treadmill walking ability. Clearly, further studies are necessary to examine the effectiveness of pharmacological agents in combination with motor training. Examining the efficacy of these pharmacological agents over the course of recovery following a SCI will be particularly important. Obviously, there is a great need to know the cellular and synaptic mechanisms through which the pharmacological agonists and antagonists may modulate locomotor function. Presently, however, there is limited availability of experimental preparations in which detailed electrophysiological studies can be performed on the adult mammalian spinal cord. CONCLUSIONS There is both supraspinal and spinal automaticity in neuromotor control systems that initiate and generate coordinated motor behavior such as standing and stepping after SCI. This automaticity is clearly reflected in the fact that the spinal cord circuitry, which has access to proprioceptive and cutaneous input, can execute standing and stepping while readily adapting to varying loads, speeds of stepping, turning, and stepping over objects.

19 PLASTICITY AFTER SCI 163 The motor patterns that the spinal cord generates after loss of supraspinal input are a function, to a great extent, of the specific sensorimotor events that occur after the injury. Training to stand improves standing ability and training to step improves stepping ability, and these behaviors are associated with motor pool specific biochemical changes. One of the biochemical consequences of a complete SCI is an upregulation of inhibitory neurotransmitter systems, and step training reverses this effect. Therefore, the physiological and biochemical state of the spinal cord will affect how it responds to any given therapeutic intervention. Ultimately, the application of any curative intervention for SCI is likely to have a higher level of success if combined with a motor training program that provides proprioceptive and cutaneous information to the spinal cord circuitry associated with the specific motor task being trained. ACKNOWLEDGMENTS These studies were supported by awards from the National Institutes of Health (1PO1NS16333; 1RO1NS40917; 1RO1NS42951), Christopher Reeve Paralysis Foundation and Consortium on Spinal Cord Injury (DA ), NASA Bioengineering Research Partnership (PAS ), and NASA GSRP (NGT ). Figure 6 is compliments of C. Liu at UCLA. The Annual Review of Neuroscience is online at LITERATURE CITED Baev KV Biological Neural Networks: Hierachical Concept of Brain Function. Boston: Birkhauser. 273 pp. Barbeau H, Fung J, Leroux A, Ladouceur M A review of the adaptability and recovery of locomotion after spinal cord injury. Prog. Brain Res. 137:9 25 Barbeau H, McCrea DA, O Donovan MJ, Rossignol S, Grill WM, Lemay MA Tapping into spinal circuits to restore motor function. Brain Res. Brain Res. Rev. 30:27 51 Barbeau H, Norman KE The effect of noradrenergic drugs on the recovery of walking after spinal cord injury. Spinal Cord 41: Barbeau H, Rossignol S Recovery of locomotion after chronic spinalization in the adult cat. Brain Res. 412:84 95 Barbeau H, Rossignol S The effects of serotonergic drugs on the locomotor pattern and on cutaneous reflexes of the adult chronic spinal cat. Brain Res. 514:55 67 Barbeau H, Rossignol S Initiation and modulation of the locomotor pattern in the adult chronic spinal cat by noradrenergic, serotonergic and dopaminergic drugs. Brain Res. 546: Bravo AB, Bigbee AJ, Roy RR, Edgerton VR, Tobin AJ, Tillakaratne NJK Gamma2 subunit of the GABA A receptor is increased in lumbar astrocytes in neonatally spinal cord transected rats. Soc. Neurosci. Abstr. 32: Calancie B, Needham-Shropshire B, Jacobs P, Willer K, Zych G, Green BA Involuntary stepping after chronic spinal cord injury. Evidence for a central rhythm generator for locomotion in man. Brain 117: Cazalets JR, Bertrand S, Sqalli-Houssaini Y, Clarac F GABAergic control of spinal

20 164 EDGERTON ET AL. locomotor networks in the neonatal rat. Ann. NY Acad. Sci. 860: Chau C, Barbeau H, Rossignol S Early locomotor training with clonidine in spinal cats. J. Neurophysiol. 79: Chau C, Giroux N, Barbeau H, Jordan L, Rossignol S Effects of intrathecal glutamatergic drugs on locomotion I. NMDA in short-term spinal cats. J. Neurophysiol. 88: Curt A, Schwab ME, Dietz V Providing the clinical basis for new interventional therapies: refined diagnosis and assessment of recovery after spinal cord injury. Spinal Cord 42:1 6 de Leon RD, Hodgson JA, Roy RR, Edgerton VR. 1998a. Locomotor capacity attributable to step training versus spontaneous recovery after spinalization in adult cats. J. Neurophysiol. 79: de Leon RD, Hodgson JA, Roy RR, Edgerton VR. 1998b. Full weight-bearing hindlimb standing following stand training in the adult spinal cat. J. Neurophysiol. 80:83 91 de Leon RD, Hodgson JA, Roy RR, Edgerton VR. 1999a. Retention of hindlimb stepping ability in adult spinal cats after the cessation of step training. J. Neurophysiol. 81:85 94 de Leon RD, Reinkensmeyer DJ, Timoszyk WK, London NJ, Roy RR, Edgerton VR Use of robotics in assessing the adaptive capacity of the rat lumbar spinal cord. Prog. Brain Res. 137: de Leon RD, Tamaki H, Hodgson JA, Roy RR, Edgerton VR. 1999b. Hindlimb locomotor and postural training modulates glycinergic inhibition in the spinal cord of the adult spinal cat. J. Neurophysiol. 82: Dietz V Spinal cord pattern generators for locomotion. Clin. Neurophysiol. 114: Dietz V, Colombo G, Jensen L, Baumgartner L Locomotor capacity of spinal cord in paraplegic patients. Ann. Neurol. 37: Dietz V, Wirz M, Colombo G, Curt A. 1998a. Locomotor capacity and recovery of spinal cord function in paraplegic patients: a clinical and electrophysiological evaluation. Electroencephalogr. Clin. Neurophysiol. 109: Dietz V, Wirz M, Curt A, Colombo G. 1998b. Locomotor pattern in paraplegic patients: training effects and recovery of spinal cord function. Spinal Cord 36: Dimitrijevic MR, Gerasimenko Y, Pinter MM Evidence for a spinal central pattern generator in humans. Ann. NY Acad. Sci. 860: Duysens J, Pearson KG Inhibition of flexor burst generation by loading ankle extensor muscles in walking cats. Brain Res. 187: Edgerton VR, de Leon RD, Harkema SJ, Hodgson JA, London N, et al. 2001a. Retraining the injured spinal cord. J. Physiol. 533:15 22 Edgerton VR, de Leon RD, Tillakaratne N, Recktenwald MR, Hodgson JA, Roy RR. 1997a. Use-dependent plasticity in spinal stepping and standing. Adv. Neurol. 72: Edgerton VR, Grillner S, Sjostrom A, Zangger P Central generation of locomotion in vertebrates. In Neural Control of Locomotion, ed. RN Herman, S Grillner, PSG Stein, DG Stuart, pp New York: Plenum Edgerton VR, Roy RR Paralysis recovery in humans and model systems. Curr. Opin. Neurobiol. 12: Edgerton VR, Roy RR, de Leon RD. 2001b. Neural darwinism in the mammalian spinal cord. In Spinal Cord Plasticity; Alterations in Reflex Function, ed. MM Patterson, JW Grau, pp Boston: Kluwer Edgerton VR, Roy RR, de Leon R, Tillakaratne N, Hodgson JA. 1997b. Does motor learning occur in the spinal cord? Neuroscientist 3: Edgerton VR, Tillakaratne NJK, Bigbee AJ, de Leon RD, Roy RR Locomotor recovery potential after spinal cord injury. In Neuro-Behavioral Determinants of Interlimb Coordination, ed. SP Swinnen, J Duysens. Boston: Kluwer. In press Eidelberg E Consequences of spinal cord

21 PLASTICITY AFTER SCI 165 lesions upon motor function, with special reference to locomotor activity. Prog. Neurobiol. 17: Fedirchuk B, Nielsen J, Petersen N, Hultborn H Pharmacologically evoked fictive motor patterns in the acutely spinalized marmoset monkey (Callithrix jacchus). Exp. Brain Res. 122: Feraboli-Lohnherr D, Barthe JY, Orsal D Serotonin-induced activation of the network for locomotion in adult spinal rats. J. Neurosci. Res. 55:87 98 Forssberg H, Grillner S, Halbertsma J. 1980a. The locomotion of the low spinal cat. I. Coordination within a hindlimb. Acta Physiol. Scand. 108: Forssberg H, Grillner S, Halbertsma J, Rossignol S. 1980b. The locomotion of the low spinal cat. II. Interlimb coordination. Acta Physiol. Scand 108: Forssberg H, Grillner S, Rossignol S Phase dependent reflex reversal during walking in chronic spinal cats. Brain Res. 85:103 7 Garraway SM, Hochman S Serotonin increases the incidence of primary afferentevoked long-term depression in rat deep dorsal horn neurons. J. Neurophysiol. 85: Gerasimenko YP, Avelev VD, Nikitin OA, Lavrov IA Initiation of locomotor activity in spinal cats by epidural stimulation of the spinal cord. Neurosci. Behav. Physiol. 33: Gerasimenko YP, Makarovskii AN, Nikitin OA Control of locomotor activity in humans and animals in the absence of supraspinal influences. Neurosci. Behav. Physiol. 32: Giroux N, Chau C, Barbeau H, Reader TA, Rossignol S Effects of intrathecal glutamatergic drugs on locomotion. II. NMDA and AP-5 in intact and late spinal cats. J. Neurophysiol. 90: Giroux N, Rossignol S, Reader TA Autoradiographic study of alpha1- and alpha2- noradrenergic and serotonin1a receptors in the spinal cord of normal and chronically transected cats. J. Comp. Neurol. 406: Gomez-Pinilla F, Ying Z, Opazo P, Roy RR, Edgerton VR Differential regulation by exercise of BDNF and NT-3 in rat spinal cord and skeletal muscle. Eur. J. Neurosci. 13: Gomez-Pinilla F, Ying Z, Roy RR, Molteni R, Edgerton VR Voluntary exercise induces a BDNF-mediated mechanism that promotes neuroplasticity. J. Neurophysiol. 88: Grillner S Interaction between central and peripheral mechanisms in the control of locomotion. Prog. Brain Res. 50: Grillner S The nervous system II. Motor control. In Handbook of Physiology Section 1. ed. VB Brooks, pp Waverly, MD: Am. Physiol. Soc. Grillner S The spinal locomotor CPG: a target after spinal cord injury. Prog. Brain Res. 137: Grillner S The motor infrastructure: from ion channels to neuronal networks. Nat. Rev. Neurosci. 4: Grillner S, Wallen P Central pattern generators for locomotion, with special reference to vertebrates. Annu. Rev. Neurosci. 8: Grillner S, Zangger P The effect of dorsal root transection on the efferent motor pattern in the cat s hindlimb during locomotion. Acta Physiol. Scand. 120: Gurfinkel VS, Levik YuS, Kazennikov OV, Selionov VA Involuntary air-stepping induced by the increase of the level of tonic readiness: evidence for CPG in human. In Brain and Movement, ed. VS Gurfinkel, YuS Levik, pp Moscow: Proc. Int. Symp. Harkema SJ Neural plasticity after human spinal cord injury: application of locomotor training to the rehabilitation of walking. Neuroscientist 7: Harkema SJ, Hurley SL, Patel UK, Requejo PS, Dobkin BH, Edgerton VR Human lumbosacral spinal cord interprets loading during stepping. J. Neurophysiol. 77:

22 166 EDGERTON ET AL. Herman R, He J, D Luzansky S, Willis W, Dilli S Spinal cord stimulation facilitates functional walking in a chronic, incomplete spinal cord injured. Spinal Cord 40:65 68 Hodgson JA, Ly L, Raven J, Roy RR, Edgerton VR, et al The effects of a unilateral corticospinal tract (CST) lesion on motor performance in Rhesus. Soc. Neurosci. Abstr. 33:71.5 Hodgson JA, Roy RR, de Leon R, Dobkin B, Edgerton VR Can the mammalian lumbar spinal cord learn a motor task? Med. Sci. Sports Exerc. 26: Ikeda H, Heinke B, Ruscheweyh R, Sandkuhler J Synaptic plasticity in spinal lamina I projection neurons that mediate hyperalgesia. Science 299: Iwahara T, Atsuta Y, Garcia-Rill E, Skinner RD Spinal cord stimulation-induced locomotion in the adult cat. Brain Res. Bull. 28: Jakeman LB, Wei P, Guan Z, Stokes BT Brain-derived neurotrophic factor stimulates hindlimb stepping and sprouting of cholinergic fibers after spinal cord injury. Exp. Neurol. 154: Jordan LM, Schmidt BJ Propriospinal neurons involved in the control of locomotion: potential targets for repair strategies? Prog. Brain Res. 137: Joynes RL, de Leon RD, Tillakaratne NJK, Roy RR, Tobin AJ, Edgerton VR Recovery of locomotion in rats after neonatal spinal transection is not attributable to growth across the lesion. Soc. Nerosci. Abstr. 29: Kiehn O, Kjaerulff O Distribution of central pattern generators for rhythmic motor outputs in the spinal cord of limbed vertebrates. Ann. NY Acad. Sci. 860: Kim D, Murray M, Simansky KJ The serotonergic 5-HT(2C) agonist m-chlorophenylpiperazine increases weight-supported locomotion without development of tolerance in rats with spinal transections. Exp. Neurol. 169: Kudo N, Nishimaru H Reorganization of locomotor activity during development in the prenatal rat. Ann. NY Acad. Sci. 860: Liu CC, de Leon RD, Guu JJ, Zdunowski SL, Melikian AH, et al Retention of a corrective kinematic response in the rat spinal cord. Soc. Neurosci. Abstr. 33: Lovely RG, Gregor RJ, Roy RR, Edgerton VR Effects of training on the recovery of full-weight-bearing stepping in the adult spinal cat. Exp. Neurol. 92: Lovely RG, Gregor RJ, Roy RR, Edgerton VR Weight-bearing hindlimb stepping in treadmill-exercised adult spinal cats. Brain Res. 514: Maegele M, Muller S, Wernig A, Edgerton VR, Harkema SJ Recruitment of spinal motor pools during voluntary movements versus stepping after human spinal cord injury. J. Neurotrauma 19: Mushahwar VK, Collins DF, Prochazka A Spinal cord microstimulation generates functional limb movements in chronically implanted cats. Exp. Neurol. 163: Nakada K, Hodgson JA, de Leon RD, Roy RR, Edgerton VR Prolonged modification of the mechanisms of the step cycle by single and repetitive mechanical stimuli in chronic spinal cats. Soc. Neurosci. Abstr. 20:1755 Orlovsky GN, Feldman AG Classification of lumbosacral neurons according to their discharge patterns during evoked locomotion neurophysiology. New York: Plenum. (From Russian) Pinter MM, Gerstenbrand F, Dimitrijevic MR Epidural electrical stimulation of posterior structures of the human lumbosacral cord: 3. Control of spasticity. Spinal Cord 38: Roy RR, Acosta L Fiber type and fiber size changes in selected thigh muscles six months after low thoracic spinal cord transection in adult cats: exercise effects. Exp. Neurol. 92: Roy RR, Hutchison DL, Pierotti DJ, Hodgson JA, Edgerton VR EMG patterns of rat ankle extensors and flexors during treadmill

23 PLASTICITY AFTER SCI 167 locomotion and swimming. J. Appl. Physiol. 70: Roy RR, Talmadge RJ, Hodgson JA, Oishi Y, Baldwin KM, Edgerton VR Differential response of fast hindlimb extensor and flexor muscles to exercise in adult spinalized cats. Muscle Nerve 22: Roy RR, Talmadge RJ, Hodgson JA, Zhong H, Baldwin KM, Edgerton VR Training effects on soleus of cats spinal cord transected (T12-13) as adults. Muscle Nerve 21:63 71 Rossignol S, Bouyer L, Barthelemy D, Langlet C, Leblond H. 2002a. Recovery of locomotion in the cat following spinal cord lesions. Brain Res. Brain Res. Rev. 40: Rossignol S, Chau C, Giroux N, Brustein E, Bouyer L, et al. 2002b. The cat model of spinal injury. Prog. Brain Res. 137: Rossignol S, Giroux N, Chau C, Marcoux J, Brustein E, Reader TA Pharmacological aids to locomotor training after spinal injury in the cat. J. Physiol. 533:65 74 Roudet C, Gimenez Ribotta M, Privat A, Feuerstein C, Savasta M Regional study of spinal alpha 2-adrenoceptor densities after intraspinal noradrenergic-rich implants on adult rats bearing complete spinal cord transection or selective chemical noradrenergic denervation. Neurosci. Lett. 208:89 92 Rygh LJ, Tjolsen A, Hole K, Svendsen F Cellular memory in spinal nociceptive circuitry. Scand. J. Psychol. 43: Schmidt BJ, Hochman S, MacLean JN NMDA receptor-mediated oscillatory properties: potential role in rhythm generation in the mammalian spinal cord. Ann. NY Acad. Sci. 860: Shik ML, Orlovsky GN Neurophysiology of locomotor automatism. Physiol. Rev. 56: Shik ML, Severin FV, Orlovsky GN Control of walking and running by electrical stimulation of the mid brain. Biophysics 11: Simonsen EB, Dyhre-Poulsen P Amplitude of the human soleus H reflex during walking and running. J. Physiol. 515(Pt. 3): Tillakaratne NJ, de Leon RD, Hoang TX, Roy RR, Edgerton VR, Tobin AJ Usedependent modulation of inhibitory capacity in the feline lumbar spinal cord. J. Neurosci. 22: Tillakaratne NJ, Mouria M, Ziv NB, Roy RR, Edgerton VR, Tobin AJ Increased expression of glutamate decarboxylase (GAD 67 ) in feline lumbar spinal cord after complete thoracic spinal cord transection. J. Neurosci. Res. 60: Timoszyk WK, de Leon RD, London N, Roy RR, Edgerton VR, Reinkensmeyer DJ The rat lumbosacral spinal cord adapts to robotic loading applied during stance. J. Neurophysiol. 88: Tuszynski MH, Peterson DA, Ray J, Baird A, Nakahara Y, Gage FH Fibroblasts genetically modified to produce nerve growth factor induce robust neuritic ingrowth after grafting to the spinal cord. Exp. Neurol. 126:1 14 Wernig A, Muller S Laufband locomotion with body weight support improved walking in persons with severe spinal cord injuries. Paraplegia 30: Wernig A, Muller S, Nanassy A, Cagol E Laufband therapy based on rules of spinal locomotion is effective in spinal cord injured persons. Eur. J. Neurosci. 7: Wirz M, Colombo G, Dietz V Long term effects of locomotor training in spinal man. J. Neurol. Neurosurg. Psychiatry 71:93 96 Wolpaw JR, Tennissen AM Activitydependent spinal cord plasticity in health and disease. Annu. Rev. Neurosci. 24: Ying Z, Roy RR, Edgerton VR, Gomez- Pinilla F Voluntary exercise increases neurotrophin-3 and its receptor TrkC in the spinal cord. Brain Res. 987:93 99

24 PLASTICITY AFTER SCI C-1 See legend on next page

25 C-2 EDGERTON ET AL. Figure 1 (A) Automaticity is a key feature of neural motor control systems. Many sources of input to the brain (blue box) and spinal cord (yellow box) are constantly processed as we accommodate to the complexities of an ever-changing physical environment. The multiple levels of sensory processing are eventually manifested in the decision as to which and how many motoneurons are activated at any given sensory state. An extensive level of sensory processing and integration by the brain and spinal cord is automatic, resulting in reasonably smart and predictable motor output (pink box) that is appropriate for the current state. This automaticity and smartness becomes more apparent after SCI (see B). (B) Spinal automaticity is highly evident after SCI. After a complete SCI, where the supraspinal connections are severed, the spinal circuitry adapts to its altered combination of inputs to facilitate motor output. The locomotion and standing patterns that can be exhibited after training illustrate not only a high level of spinal automaticity but also the spinal cord s capacity to learn and perform motor tasks. The relative importance of the spinal circuitry and its sensory input is much greater than normal (see A) as illustrated by the thicker arrows into and out of the spinal cord, whereas the direct supraspinal influence on locomotion is greatly reduced. The gray box lists various therapeutic strategies by which functional motor recovery may be enhanced after SCI.

26 Annual Review of Neuroscience Volume 27, 2004 CONTENTS THE AMYGDALA MODULATES THE CONSOLIDATION OF MEMORIES OF EMOTIONALLY AROUSING EXPERIENCES, James L. McGaugh 1 CONTROL OF CENTRAL SYNAPTIC SPECIFICITY IN INSECT SENSORY NEURONS, Jonathan M. Blagburn and Jonathan P. Bacon 29 SENSORY SIGNALS IN NEURAL POPULATIONS UNDERLYING TACTILE PERCEPTION AND MANIPULATION, Antony W. Goodwin and Heather E. Wheat 53 E PLURIBUS UNUM, EX UNO PLURA: QUANTITATIVE AND SINGLE-GENE PERSPECTIVES ON THE STUDY OF BEHAVIOR, Ralph J. Greenspan 79 DESENSITIZATION OF G PROTEIN COUPLED RECEPTORS AND NEURONAL FUNCTIONS, Raul R. Gainetdinov, Richard T. Premont, Laura M. Bohn, Robert J. Lefkowitz, and Marc G. Caron 107 PLASTICITY OF THE SPINAL NEURAL CIRCUITRY AFTER INJURY, V. Reggie Edgerton, Niranjala J.K. Tillakaratne, Allison J. Bigbee, Ray D. de Leon, and Roland R. Roy 145 THE MIRROR-NEURON SYSTEM, Giacomo Rizzolatti and Laila Craighero 169 GENETIC APPROACHES TO THE STUDY OF ANXIETY, Joshua A. Gordon and René Hen 193 UBIQUITIN-DEPENDENT REGULATION OF THE SYNAPSE, Aaron DiAntonio and Linda Hicke 223 CELLULAR MECHANISMS OF NEURONAL POPULATION OSCILLATIONS IN THE HIPPOCAMPUS IN VITRO, Roger D. Traub, Andrea Bibbig, Fiona E.N. LeBeau, Eberhard H. Buhl, and Miles A. Whittington 247 THE MEDIAL TEMPORAL LOBE, Larry R. Squire, Craig E.L. Stark, and Robert E. Clark 279 THE NEURAL BASIS OF TEMPORAL PROCESSING, Michael D. Mauk and Dean V. Buonomano 307 THE NOGO SIGNALING PATHWAY FOR REGENERATION BLOCK, Zhigang He and Vuk Koprivica 341 MAPS IN THE BRAIN: WHAT CAN WE LEARN FROM THEM? Dmitri B. Chklovskii and Alexei A. Koulakov 369 v