Plasticity of Functional Connectivity in the Adult Spinal Cord

Author(s): L. L. Cai, G. Courtine, A. J. Fong, J. W. Burdick, R. R. Roy and V. R. Edgerton

Source: Philosophical Transactions: Biological Sciences, Vol. 361, No. 1473, The Regenerating
Brain (Sep. 29, 2006), pp. 1635-1646
Published by: The Royal Society
Stable URL: http://www.jstor.org/stable/20209754 .
Accessed: 17/01/2015 03:16

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].

The Royal Society is collaborating with JSTOR to digitize, preserve and extend access to Philosophical
Transactions: Biological Sciences.

http://www.jstor.org

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
PHILOSOPHICAL
TRANSACTIONS
Phil. Trans. R. Soc. B (2006) 361, 1635-1646
THEROYAL Hf\
doi: 10.1098/rstb.2006.1884
SOCIETY JUJ Published online 11 August 2006

Plasticity of functional connectivity in the adult

spinal cord
L. L. Cai1, G. Courtine4, A. J. Fong1, J. W. Burdick1'2,
R. R. Roy3 and V. R. Edgerton3'4'5'*
1
Bioengineering Option, Engineering Option, California Institute of Technology,
2Mechanical
Pasadena, CA 91125, USA
4
3Brain Research Institute, Department of Physiological Science, and 5Department ofNeurobiology,
University of California, Los Angeles, CA 90095, USA
This paper emphasizes several characteristics of the neural control of locomotion that provide
opportunities for developing strategies to maximize the recovery of postural and locomotor functions
after a spinal cord injury (SCI). The major points of this paper are: (i) the circuitry that controls
standing and stepping is extremely malleable and reflects a continuously varying combination of
neurons that are activated when executing stereotypical movements; (ii) the connectivity between
neurons is more accurately perceived as a functional rather than as an anatomical phenomenon;
(iii) the functional connectivity that controls standing and stepping reflects the physiological state of a
given assembly of synapses, where the probability of these synaptic events is not deterministic;
(iv) rather, this probability can be modulated by other factors such as pharmacological agents,
epidural stimulation and/or motor training; (v) the variability observed in the kinematics of
consecutive steps reflects a fundamental feature of the neural control system and (vi) machine
learning theories elucidate the need to accommodate variability in developing strategies designed to
enhance motor performance by motor training using robotic devices after an SCI.

Keywords: spinal cord injury; rehabilitation; robotics; locomotion; standing; neural control systems

1. INTRODUCTION sensory and motor circuits that are involved in the

The title of this
paper may induce a myriad of generation of posture and locomotion. A measure of
perceptions, most
of which will imply physiological functional connectivity, in the context of how we are
mechanisms related to how the adaptation of neural using this term, is the probability of a specific set of
events within the central nervous system responds to a neurons being activated for a given physiological state.
spinal cord injury (SCI). Clearly, after an injury of any Many correlations have been drawn between ana
part of the neuromuscular system, there are changes in tomical connections and functional recovery post-SCI
the connectivity of those sensorimotor circuits that (Hase et al 2002; Lee et al 2004). However, the
generate a motor task. Changes also occur during the variability in normal stepping, even under well
subsequent adaptations that follow the injury. In this controlled conditions, demonstrates the versatility
paper, emphasis will be on the concept of functional and complexity in the activation of the associated
rather than anatomical connectivity within the spinal spinal circuitry. We propose that as the physiological
cord. The term functional connectivity will be used to states change, the continuous adaptation in functional
indicate that the likelihood of a given neuron being connectivity brings about routine variability in the
activated is dependent on its physiological state of activation patterns during repetitive movements, such
'readiness' rather than on the existence of an as stepping. As a result, no two steps are generated by
merely
anatomical connection. This emphasis is not to imply the same combination and sequence of neuronal
that changes in the actual number of activation. As the limb trajectory varies from step to
synaptic
connections cannot or do not occur in response to step, the precise pattern of activation of the motor pools
SCI. In fact, there is good evidence for the presence involved must also vary (figure 1). This variation is
of such adaptations, and these can be reflected in the electromyographic (EMG) signals from
changes
associated with an improvement inmotor performance normal (Courtine et al 2005a) and complete spinal
capacity following an SCI (Bregman et al 1997, 2002;
animals (Lovely et al 1990; Edgerton et al 1992)
Raineteau & Schwab this paper will
2001). Instead, stepping on a treadmill. Thus, even within the robust
focus on the importance of rapid, sometimes and size principle of recruitment of motor neurons (Burke &

persistent, changes in functional connectivity between Edgerton 1975; Henneman & Mendell 1981), there
a given combination of spatially and linked remains a significant level of variability in the exact
temporally
combination and order of motor neurons activated
within a given motor pool, and certainly across
*Author for correspondence ([email protected]). synergistic motor pools (Cope & Sokoloff 1999), to
One contribution of 13 to a Theme Issue 'The regenerating brain'. generate a specific movement.

1635 This journal is? 2006 The Royal Society

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 1636 L. L. Cai and others Spinal cord plasticity after injury

SOL completing a successful step (Lovely et al 1986;

(a) Barbeau & Rossignol 1987; de Leon et al. 19986). It
14
force appears from the results of virtually all the studies

1 (N) 1.6]
motor after an SCI that the benefits
0 involving training
of step training can be manifested as an increased
U"
EMG probability of generating a successful step. At the
(mV) _j 7 systems level, a number of motor training-induced
9.8 biochemical and electrophysiological changes in the
force
spinal cord are associated with improved motor
1W 0 performance after an SCI (de Leon et al 1999;
? EMGL51 Tillakaratne et al. 2000, 2002; Cote et al. 2003;
(mV) Cote & Gossard 2004).
-1.2

MG

(b) 2.5 2. SOME BIOCHEMICALAND ELECTROPHYSIO

force LOGICALCHANGES ASSOCIATED WITH
- (N) IMPROVEDMOTOR PERFORMANCE INSPINAL
g.S 0
0 4 ANIMALS
& EMG
Prior research has identified a number of biochemical
and physiological changes in the spinal cord after a
complete thoracic spinal cord transection in response
to training of a specific motor task. Many of these
changes have been reviewed recently (Edgerton et al.
2004). Briefly, the biochemical changes generally
reflect an upregulation of both the glycinergic and
2000 acid neurotrans
gamma-aminobutyric (GABA)ergic
time (ms)
mitter systems within the lumbosacral spinal cord.
Figure 1. Force and EMG records from the (a) soleus (sol) The biochemical indicators consist of an increased
and (?>) medial gastrocnemius (MG) muscles of a normal, number of glycinergic receptors (Edgerton et al.
intact cat and an adult spinal cat during stepping on a
2001a), an increased responsiveness to strychnine
treadmill at 0.8ms"1. Bold bars indicate the period of an agent that blocks the glycinergic
administration,
contralateral with the cat
support. Compared normal, spinal
receptor (de Leon et al 1999), an increased level
exhibited a longer cycle period, a steeper decline in force
of glutamic acid decarboxylase (GAD67; Tillakaratne
beginning at mid-support, a delay in the onset of flexion at the
et al 2000) and improved locomotion when blocking
ankle (Fa), lower peak EMG forces, and clonus in the EMG
and force records of both muscles. PC, paw contact (taken
GABAergic inhibition with biccuculine (Robinson &
from et al. Goldberger 1986a,?>). However, an important obser
Lovely 1990).
vation is that the increased level of inhibition in the

The source of this variability in stepping is spinal cord after an SCI can be countered by motor
training (Edgerton et al 2001a; Tillakaratne et al
undoubtedly derived from both supraspinal as well as
2002; Edgerton et al. 2004).
spinal neuronal networks. It is also reasonably obvious
It appears that repetition of a specific motor task
that the variability in limb kinematics will be greater
can modulate the level of persistent inhibition
following an SCI, as recovery of stepping occurs either in the neural networks that normally
or as a result
present
spontaneously of motor training. After an
generate the motor task. These effects have been
SCI, however, the variability in stepping is significantly
demonstrated in spinal animals that have been
reduced by motor training (de Leon et al. 19986). We trained to step (de Leon et al 19986) or stand (de
have proposed that this variability reflects the intrinsic et al as
Leon 1998a). However, it remains unclear
probability of a given assembly of neurons being to how versus neural
specific excitatory inhibitory
activated at any given time (Edgerton et al 20016).
pathways are modulated by repetitive use. Repetitive
Thus, the underlying explanation for the presence of use of the extensor musculature may downregulate
variability in the limb kinematics during stepping under the associated with extensor motor
GAD67 neurons,
normal conditions is that whether or not an assembly and simultaneously enhance the levels of GAD67 by
of neurons is activated is not deterministic at any the inhibition of flexor motor neurons
increasing
given instant. et al 2002). A limitation of these
(Tillakaratne
After an SCI, the probability that an appropriate observations is that linking the level of excitation
combination of neurons is activated in the appro versus inhibition of specific neural to
pathways
priate sequence is markedly altered. During the within the different neurotrans
specific adaptations
'reorganization' of the spinal circuitry following mitter systems has not been possible to date.
SCI, these probabilities can become lower or higher Furthermore, there has been relatively little identifi
depending, to a large degree, on the frequency with cation of the receptor subtypes that may be
which the sensorimotor circuits experience specific associated with the observed level of behavioural
patterns of activity. For example, the repetitive performance. These data are further limited as it is
performance of a motor task, such as stepping, over not certain whether the observed biochemical
a period of weeks increases the probability of changes are simply correlated with the changes in

Phil. Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 Spinal cord plasticity after injury L. L. Cai and others 1637

motor performance as opposed to there being a 3. GENERAL CONTROL DEMANDS: HIERARCHI

cause and effect relationship. Further studies are CALLY DESIGNED NETWORKS
needed to address these issues. Several observations demonstrate that supraspinal
Electrophysiological changes have also been control can be, and probably often is, relatively
observed in chronic spinal animals, and there is strong non-specific. Supraspinal input can instruct the spinal
evidence that the efficacy of selected neuromotor cord to walk by providing a relatively non-specific
pathways can be modified by repetitive training of a tonic input, leaving the detailed decisions of which
motor task. For example, there is an improved neuronal systems have to be activated at the spinal
coordination of motor pools controlling the hindlimb level. Stepping can be induced in decerebrated
musculature in animals via tonic stimulation of the mesencephalic
following step training spinal animals,
as shown by EMG bursting patterns (Lovely et al
locomotor region (Shik et al 1966). Fictive loco
motion can be generated via stimulation of the dorsal
1990). Likewise, the step training greatly improves the
roots of the spinal cord (Sjostrom & Zangger
transmission in polysynaptic excitatory group I load 1976).
et al that locomotor Locomotion can be induced pharmacologically in
pathways (Cote 2003) convey
drive to extensor motor neurons, and thus could complete spinal animals (Rossignol & Barbeau 1993;
contribute to of weight de Leon et al 1999; Antri et al 2002; Orsal et al
improved recovery bearing
in animals. The mean amplitude 2002), and chronic complete, low thoracic spinal
during stepping spinal
animals can be trained to step even without any
of excitatory post-synaptic potentials has similarly been
reported to increase in response to activation of skin pharmacological enhancement (Lovely et al 1986;
Barbeau & Rossignol 1987; de Leon et al 19986).
sensory receptors located under the paw of chronic
cats to step (Cote & Another illustration of the non-specific nature of the
spinal that have been trained
also indicate that signals (in this case, from the periphery) that generate
Gossard 2004). Recent observations
walking is the locomotor-like movements that can be
improved stepping following training in complete
elicited in humans in a recumbent position by
spinal rats correlates with the peak amplitude of the
applying non-specific tonic vibration to the relaxed
segmental excitatory post-synaptic and action
leg musculature (Gurfinkel et al 1998). These widely
potentials and level of afterhyperpolarization depth of
different manipulations represent an impressive
the motor neurons recruited during locomotor activity
array of different techniques that change the physio
(Petruska et al 2004). state a
logical of the spinal cord, all having
Based on the results of these electrophysiological
remarkably similar effect, i.e. they induce or improve
studies, it appears that in chronic spinal animals, a
number of spinal neural can respond stepping ability.
pathways The generality of the supraspinal, and even spinal,
specifically to step training. Accordingly, it is likely
commands is also apparent from the strong inter
that after repetitive exposure of the spinal cord to a
relationships among the kinematics of multiple joints
given motor task, significant alterations may occur in within and across limbs during locomotion in intact, as
the transmission of many, if not all, sensorimotor
well as complete and incomplete SCI, animals
pathways caudal to the lesion. Such task-dependent
(figure 2). This stereotypical output implies a close
functional plasticity of the spinal motor infrastructure
link between the neuronal systems controlling each of
could, in turn, contribute to the decrease in the
these joints and all the musculatures associated with
intrinsic variability in performing a motor task (de
their dynamics. Such a high intrinsic coupling in the
Leon et al 19986), i.e. there would be an increase in the
generation of limb oscillation also simplifies the details
probability of activating specific motor pathways, and
required by the brain to generate a complex motor task,
therefore specific functional sets of neurons, to
such as stepping at a range of speeds (Bianchi et al
accomplish the required task (Edgerton et al 20016).
1998), grades (de la Torre & Goldsmith 1990) and
These results suggest an improved efficacy of the directions (Courtine & Schieppati 2004). Further
interneurons that are responsible for coordinating this the of
more, stereotypical output supports concept
motor la interneurons that
pools, e.g. provide recipro
automaticity in the control of locomotion (Orlovsky &
cal inhibition between antagonistic motor pools. Direct Feldman 1972). Briefly, automaticity is the ability to
measurements of decreases followed by increases in a range of motor such as and
generate tasks, stepping
the excitability of synapses associated with la inter in response to highly predictable ensembles
standing,
neurons in response to spinal cord transection have of sensory stimuli from the periphery and motor
been observed (Valero-Cabre et al 2004a,6). On the commands from the brain. Considerable automaticity
other hand, significant increases in the excitability of remains within the spinal cord in the absence of
lumbar monosynaptic reflexes have been reported supraspinal input.
(Thompson et al 1998). One source of such automaticity is found in the
Thus, it seems reasonable to hypothesize that the organization of the spinal circuits generating the motor
observed biochemical and electrophysiological adap patterns for walking. Stimulation of such neural
tations of spinal animals, associated with step circuits, often referred to as central pattern generators
training, are not due to an induction of a specific (CPGs), produces rhythmic alternating flexor and
set of synaptic events required for the acquisition of extensor activities in several vertebrate models, e.g.
stepping ability. Rather, these adaptations result lamprey eels, neonatal rats or adult cats (Arshavsky
from the increased probability of the neural et al 1997; Grillner 2002; MacKay-Lyons 2002) that
circuitries within the spinal cord of generating a mimic locomotion. However, the structural organiz
successful step. ations of these neural circuitries inmammals are not yet

Phil. Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 1638 L. L. Cai and others Spinal cord plasticity after injury

(a) humans (b) monkeys (c)

60i1?lllilk J^) f^-$^???^ ? ? * t *a

?
* ? L^m^S 1 I ?ill
3Hk mmW I III!
d
x^- ^pf^> v#
^> 1
60J
2 6 12
^ PRE POST
(weeks)
Figure 2. Coupling in the generation of limb movements during walking in humans and monkeys, (a) When plotted in a three
dimensional space, the angular oscillation of thigh, shank and foot segment with respect to the direction of gravity, i.e. elevation

angles, covaries close to a plane, both during (a) human and (6) monkey locomotion. The gait loop evolves in the
counterclockwise direction. The onset of stance (St) and swing (Sw) are indicated, (c) The degree of coupling among limb
movements is evaluated by
applying a principal component (PC) analysis on the elevation angles of hindlimb segments (thigh,
shank and
foot). Mean
(s.d.) values of the variance explained by the first PC during treadmill locomotion performed pre-lesion
(PRE) and 1, 2, 6 and 12 weeks after a unilateral lesion to the thoracic dorsolateral column (POST) is shown for three monkeys.
Asterisks indicate significant difference between pre- and post-lesion values. The high variance accounted for by the first PC
reflects the high degree of coupling in the neuronal systems that generate the oscillation of the limbs during stepping both in
intact and spinal cord-injured animals (adapted from Courtine et al. 2005a).

fully understood. Nonetheless, even within the most levels observed pre-lesion (figure 3) does not imply
automated action from CPGs, automatic adjustments re-establishment of pre-lesion muscle synergies, but
can be made by varying levels of hierarchical control. instead reflects novel activation patterns of inter
For example, extensor or flexor muscle activity can neurons and motor pools that may be possible as a
be altered independently during fictive locomotion result of the hierarchal features noted earlier (Courtine
without affecting the ongoing locomotor rhythm et al. 20056). It has also been found that new motor
(Lafreniere-Roula & McCrea 2005). This observation patterns underlie the learning of foot kinematics that
suggests that the spinal-generated rhythmical input are similar to those of non-disabled individuals during
drives multiple pattern formation modules, and treadmill stepping following training in humans with an
activities of these modules can be modified by other incomplete or complete SCI (Wernig et al 1995;
sources, e.g. sensory or supraspinal inputs. Such Grasso et al. 2004). Such findings provide evidence
hierarchical control would introduce both stereotypical that successful stepping, as defined kinematically, can
and ongoing adjustments (variability) in the motor be achieved through activation of a variety of spinal
output to adjust locomotor kinematics. This hierarchal motor neurons, and that there is no fixed locomotor
control can range from volitional circuits to extreme circuitry for the generation of stepping in mammals,
automaticity, i.e. CPGs. Although there is still no direct including primates.
evidence for the existence of locomotor spinal circuits Besides being able to accommodate the control that
that display CPG properties in humans, Dimitrijevic can be exerted by the brain on specific interneuronal
et al (1998) reported that non-patterned electrical assemblies or motor pools, even more direct neural

stimulation of the posterior structures of the lumbar connections must be present to control specific muscle
spinal cord in subjects with complete, long-term SCI units or combinations of units (figure 4). It is generally
can induce rhythmic, alternating stance and swing accepted, at least in primates, that there are direct
phases of the lower limbs. projections from the corticospinal tract (CST) to spinal
This predictability of a Stereotypie stepping pattern motor neurons (cortico-motoneuronal connections),

may seem contradictory to the concept noted earlier although a small portion of the corticospinal projections
that variability in stepping reflects an important feature actually represent a direct target to motor neurons

of the neural control system. However, it should be (Lacroix et al 2004). The cortical projection to the
recalled that spinal interneurons receive input from spinal cord can be an important source of modulation in
supraspinal, as well as from spinal, sources, and from the production of skilled locomotor movements, such
the periphery. Nevertheless, given the apparent hier as stepping over an obstacle or the precise positioning
archical organization of the neuronal systems that of the paw during walking in cats and primates
generate motor patterns and interpret sensory infor (Lawrence & Kuypers 1968; Georgopoulos & Grillner
mation as implied earlier, there are multiple com 1989; Drew et al 2002; Courtine et al 2005a,b).
binations and levels of neuronal control systems that Theoretically, corticospinal input to subsets of inter
underlie the variability observed during stepping while neurons that control specific combinations of muscle
simultaneously maintaining a very high probability of units in specific motor pools (Fetz et al 2000; Drew et al.
success from step to step (figure 3). 2002; Lemon et al. 2004) could allow precise voluntary
It is worth noting that recovery of locomotion activation and frequency control of groups of motor
kinematics after an incomplete SCI in the monkey to units (Kuypers 1978).

Phil Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 Spinal cord plasticity after injury L. L. Cai and others 1639

ib) 1.5 o

I
^ ^ y supraspinal
0.5 J sol
* O D Q
2 *

1 11
'spinal
MG controller'
0
1.5

0.5 i
VL motoneuron
0
x> 1.5
"O
N muscle unit
3 1 1
FHL
0.5 Figure 4. Cartoon depicting several features of the sensor
6 * * imotor control of movement. The cartoon illustrates the
FDL
4 possibility of a supraspinal control centre with neurons
projecting to control neurons
level controllers' of
2i ('spinal
i movements of differing complexities) that would project to a
25 50 75 100 0
group of synergistic motor pools, muscles and muscle units.
1.5i
cycle duration (%) In cases illustrated by the projection of neuron a or neuron b,
1 specific control of a small group of motor units be
! might
unnecessary in executing a generalized motor programme to
pre-lesion 0.5 JTA control The numbers 1-5 denote five muscle units.
stepping.
post-lesion 1wk
1 12 The dots embedded in the triangles represent individual
post-lesion 12 wks
neurons. Activation of neuron a would result in muscle units
PRE POST
1-4 being recruited. Neuron b would recruit muscle units
(weeks)
2-5, whereas neuron c would recruit only muscle unit 5. On
3. (a) Mean the other hand, there can be even more selective control of
Figure (s.d.) waveforms of each joint angle for
the hindlimb to the lesion side recorded motor units as illustrated with neuron c. At least for some
ipsilateral during
treadmill muscle in some
locomotion (0.45 m s-1) before (pre-lesion) and 1 groups species, there may be direct
and 12 weeks after a unilateral of supraspinal connections to some motor pools as well as the
(post-lesion) interruption
the lateral CST in the thoracic cord of adult Rhesus more generalized command neurons that exert more
spinal general
= control motor Two sets of divergent
monkeys (n 2). The horizontal bars at the bottom indicate signals among pools.
the mean are illustrated to point out the
(s.d.) value of the stance phase duration. (?>)Mean triangles flexibility in
(s.d.) values of EMG burst for all recorded muscles. modulating the set of muscles that may be recruited for a
integrals
Sol, soleus; MG, medial gastrocnemius; VL, vastus lateralis; given movement. One can also view the upright triangles in
FHL, flexor hallucis extensor the reverse direction (see arrows labelled
longus; EDL, digitorum projecting upward,
tibialis anterior. Values are normalized to the as d), a single
longus; TA, symbolizing sensory receptor projecting
pre-lesion baseline (dashed as the mean rostrally and diverging markedly, thus a single
lines) computed illustrating
value of muscle locomotion. sensory receptor that could provide excitatory or
activity during pre-lesion inhibitory
Asterisks indicate difference between and input to a large number of neurons within the spinal cord.
significant pre-
post-lesion values from Courtine et al 20056). This sensory information, in turn, may further or
(adapted diverge
even converge to specific locations. The
supraspinal diverging
4. HIERARCHICALCOMMANDCOMBINEDWITH circuits that enable different levels of control of multiple
SMART SENSORY CONTROL muscles also provide a means of detailed conscious control of
fine movements, while also mechanisms for
The progression from a supraspinal motor command to providing
executing more general and even when
the detailed control of hundreds of thousands of muscle predictable tasks,
they are considerably complex.
fibres reflects a phenomenal rostral-to-caudal ana
tomical and physiological divergence. Although signals from multiple receptors merge and project at
convergence of inputs from cortical cell populations some level to common, as well as
unique, combinations
onto motor neuron assemblies there is also a
exists, of neurons within the spinal cord. Every unique
remarkable divergence in how the supraspinal drive can combination of neurons in turn,
can, readily recognize
affect various interneuronal circuits (figure 4). Simi and very specific sensory patterns
complex that can
larly, projections of the signals from specific sensory trigger the appropriate motor responses for a given
receptors in the periphery to the spinal cord networks
pattern of sensory information. In other words, a given
and brain are highly divergent. For example, a single
pattern of sensory information provides very specific
muscle spindle alerts thousands of neurons within the and recognizable information to the neurons that
spinal cord (and even more in the brain) that a signal
eventually generate the appropriate motor response to
related to the physical environment of the mechan that given ensemble of sensory information. To what
oreceptor has been generated (Scott & Mendell 1976). degree this extensive divergent and convergent infor
It is inevitable that all the sensory information from the mation from the periphery is processed and integrated
periphery project to the networks of neurons within within the spinal networks to relaying this
prior
the spinal cord. At the same time, it is apparent that the information to the brain is unknown, but it is readily

Phil Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 1640 L. L. Cai and others Spinal cord plasticity after injury

apparent that these complex patterns of sensory input Other observations support the importance of the
provide a continuous stream of critical information for interaction between CPG and sensory processing.
the ongoing control of specific motor responses, such as Results similar to those described for strychnine earlier
stepping and standing. were observed when the dorsum of the lumbosacral
These observations are not meant to imply that the spinal cord of complete mid-thoracic spinal rats
neural circuits within the spinal cord segments have a (Ichiyama et al. 2005) and cats (Gerasimenko et al
lesser role than the brain circuits in controlling 2003, 2005) was stimulated via epidural electrodes. In
stepping. For example, the ability of the spinal circuitry this case, a tonic general stimulation of modest
to generate rhythmic motor patterns that mimic intensity applied to the dorsum of the spinal cord did
locomotion without any sensory input, i.e. fictive not induce any rhythmic, step-like motion. When the
locomotion, is clear (Grillner 2003). This fictive central hindlimbs were placed on a moving treadmill belt,
pattern generation, however, cannot make any adjust however, the animals stepped at a rate consistent with
ments to changes in its environment and, therefore, the speed of the treadmill belt. Previous experiments
there are no mechanisms to change stepping frequency, have also demonstrated that complete spinal cats
modulate the appropriate level of load bearing, or receiving tonic electrical stimulation of the dorsum of
adjust to any, even slight, perturbation. In fact, not only the lumbar spinal cord can step backwards when their
does the afferent input provide the spinal locomotor hindlimbs are placed on a treadmill belt moving
circuits with information related to unexpected events, forward (Gerasimenko et al 2003). These data
but the ongoing sensory flux also contributes substan demonstrate that detailed complex signals that drive
tially to the activation of motor neurons, even under motor pools in a highly coordinated fashion can be
normal walking conditions (Pearson 2004). On the derived from very general patterns of stimuli to the
other hand, this central pattern generation, combined lumbosacral spinal cord. Furthermore, these experi
with its massive online sensory information processing ments clearly indicate that the sensory information
capability, can effectively generate motor tasks such as provided to the spinal cord essentially defines the
stepping and standing without input from the brain type of motor task that will be performed, as well
(Edgerton et al 2004). as the characteristics of the motor pattern associated
with the task.

5. SENSORY MODULATIONOF MOTOR TASKS

What is the evidence that the sensory information 6. IMPLICATIONSOF SYNAESTHESIA FOR
derived from the limbs during posture and locomotion REHABILITATION
represent a critical and primary influence on motor Synaesthesia is the merging of different modes of
control in complete spinal animals? Several experi sensation received by the nervous system. Each mode
ments demonstrate that sensory information can define of sensation, e.g. hearing, seeing or
touching, is
most details of all postural and locomotor movements. generally thought to be very closely linked with specific
Administration of strychnine, which blocks glycinergic types of sensory receptors providing information to
inhibition, at a dose that did not generate spontaneous areas of the brain that have the capability to process
rhythmic motion of the hindlimbs, facilitated consist sound, light or mechanical perturbation, respectively.
ent, full weight-bearing treadmill stepping of the There are many examples of how sensory modes can be
hindlimbs in chronic complete spinal cats that other merged or exchanged with respect to a predictable
wise could not step (de Leon et al 1999). In addition, perception generated by a sensor. For example,
the rate of stepping was modulated to accommodate Cytowic (2002) described a subject who was born
the speed of the treadmill belt. These results demon blind, but later regained vision. After the vision was
strate that the spinal cord was not induced to generate restored, this individual had difficulty in seeing an
rhythmic activity and stepping by strychnine itself, but object without touching itwith his hands. For example,
that strychnine changed the physiological state of the when he saw a gorilla at a zoo, he could not understand
spinal neural circuits so that the sensory information its posture and movements until he had felt a statue of a
could be processed and transformed with sufficient gorilla. There are also impressive examples of utilizing
accuracy to control locomotion over a range of speeds this synaesthetic capability to rehabilitate individuals
and levels of loading. Similar observations have been who had their vestibular system destroyed by medi
made after the administration of quipazine, a seroto cation. Individuals who have extreme difficulty in
nergic agonist, to mice having a complete SCI (Fong standing and walking as a result of a pharmacologically
et al 2003, 2005). When combined, these results induced loss of vestibular function can rapidly regain
clearly demonstrate that the sensory input associated excellent control by substituting the vestibular infor
with standing and stepping generates successful and mation with the output from an accelerometer placed
remarkably adaptive control of posture and locomotion on the head. In these cases, the electrical output from
in the absence of supraspinal input. Under these the accelerometer was passed via a wire leading to the
conditions, this adaptive control cannot be solely surface of the tongue (Tyler et al 2003). In some way,
attributed to central pattern generation, i.e. repetitive the subject's tongue was able to 'calibrate' the

cycles of flexion and extension. A very important accelerometer output with visual and, presumably,
additional feature of the neural circuitry that generates head, neck, trunk and lower limb proprioceptive
these patterns is its ability to interpret the sensory input signals, functionally merging the information so that
in a manner that becomes meaningful to the success of virtually normal posture and locomotion could be
the hindlimbs in responding to its environment. sustained. Furthermore, it is interesting that once the

Phil Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 Spinal cord plasticity after injury L. L. Cai and others 1641

accelerometer device was removed, the renewed resulting output is minimized (Zhou et al 1996;
control of posture and movement was maintained for figure 5 a). Although such controllers have proven to
days or even weeks. Essentially, the brain of this patient be versatile and reliable in the engineering world, it is
was able to substitute electrical signals derived from an unlikely that biological neuromuscular systems are
accelerometer and integrate this information into the controlled in such a manner. Firstly, such a control
circuitry that coordinates the musculature of the head, system cannot adapt to changes in the environment and
neck, trunk and lower limbs that performs postural and can become unstable quickly when there are large
locomotor tasks. disturbances. Secondly, such a control system lacks the
With respect to the topics of this paper, the concept ability to learn to improve its performance from past
of synaesthesia may be important in several ways when experience. In contrast, the vertebrate spinal cord
developing strategies to recover sensorimotor function. neural circuitry is capable of adjusting to disturbances
Perhaps the most important point from these obser (Barbeau et al. 2002) and learning from repetitive
vations on synaesthesia is the degree to which the brain training (Edgerton et al 1992; de Leon et al 1998a,b),
can reorganize its function, even in individuals without even in the absence of supraspinal control.
any detectable neural dysfunction. This raises the Another robust property of the normal neural
question as to what extent we can learn to substitute control system of locomotion is variability. Even with
one sensory mode for another in facilitating recovery of this variability, the success of hundreds or even
function after an SCI. Following a severe, functionally thousands of steps can be predicted in the uninjured
incomplete SCI, for example, to what extent can the individual under normal circumstances. Following an
brain reorganize itself to use the small number of intact SCI, this variability in both the activation patterns and
fibres to functionally project signals to the spinal cord the resulting kinematics of the hindlimbs increases, and
below the lesion? In other words, can a residual source the probability of generating consecutive successful
of control from the brain be modified to control a steps will be quite low and, in many cases, near zero.
function that is different from its normal action? Step training reduces the variability in the kinematics of
A second important point that can be derived from the limb motions (de Leon et al 19986). Presumably,
these examples of synaesthesia is that two modes of increasing the occurrence of a given pattern of sensory
sensory information can be substituted, or at least information associated with load-bearing stepping
merged, to improve sensorimotor function. increases the probability of pattern recognition by the
Another example of functional sensorimotor neural networks that are linked to the sensory patterns.
reorganization after an injury is the perception of the In addition, frequent occurrence increases the prob
phantom limb, with a subject sensing the presence, and ability of generating a predictable kinematic pattern
even the touch, of an arm that has been amputated whenever this sensory pattern is recognized. In essence,
(Kuiken et al 2004). Subjects who have had an arm the likelihood of a given set of neurons being activated
amputated can learn to control prosthetic devices using in a given condition may change from near randomness
the EMG signals derived from intact muscles of the to one that is highly predictable, reflecting properties
shoulder or from shoulder muscles that have been that are typical features of learning systems.
re-innervated with nerve branches which originally Most machine learning theories of learning systems
innervated muscles controlling hand and wrist move are developed based on 'supervised learning', a process
ments. Interestingly, touching the skin overlying these of learning from examples provided by a knowledgeable
re-innervated muscles gives the subject the sensation of instructor, where the input and output examples are
touching the skin overlying the hand or wrist, i.e. the provided as a classified pair (Anderson et al 1983).
region that it normally innervates. Although this type of learning is significant in knowl
All these observations indicate that the potential for edge acquisition, this alone is not adequate to explain
reorganization of sensorimotor function after an SCI locomotor learning. In a neuromuscular system, most
has not been fully realized as a rehabilitative strategy. of the learning processes are unsupervised and involve
Combining this potential for plasticity with new skill refinement. Although the idea has been around for
technologies, such as virtual reality and smart robotic decades, a relatively new set of learning theories called
devices, seems to be a feasible and logical direction for 'reinforcement learning' has been developed in recent
future efforts in enhancing recovery of sensorimotor years to address this problem.
function following a wide range of neuromotor In contrast to supervised learning, reinforcement
disorders. For example, robotic devices can be used learning emphasizes learning feedback that evaluates
to provide more precise and versatile training to SCI performance without providing a standard of correct
subjects (mentioned later). ness in the form of behavioural targets, i.e. reinforce
ment learning gives an index of how well the system
performed relative to its previous trials without giving
7. AN ENGINEERINGPERSPECTIVE ON THE any indication of the correct response (Barto 1994).
IMPORTANCEOF VARIABILITY INA CONTROL Therefore, to maximize the reward, the reinforcement
SYSTEM learning paradigm requires the system to actively try
The automaticity in the control of locomotion as well as alternatives, evaluate the results, and then use a
the importance of sensory feedback in neuromuscular selection mechanism to guide the behaviour towards
systems (mentioned earlier) resemble that of a simple the best alternative. The fundamental process is
mechanical control system, where the goal is to use analogous to 'trial and error'. However, the search is
feedback to guide the performance of amovement such not random or undirected. Instead, the system takes
that the error between the command input and the into account results acquired from previous trials to

Phil. Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 1642 L. L. Cai and others Spinal cord plasticity after injury

(a)

disturbance

command
controller plant ->
output
input ^o

feedback

(b) disturbance

command
-
input output

Figure (a) Block

5. diagram for a simple mechanical controller. In neuromuscular systems, the controller would be the motor
neurons, the plant would be the muscles and the sensor would be all the proprioceptive feedbacks to the motor neurons. The
?
information provided by the sensor is a negative feedback (denoted by the sign) and is used to minimize the error between
the output of the plant and the command input, a positive input (denoted by the + sign), the command input. In controls, the
disturbance generally refers to unmodelled dynamics of the plant. However, in neuromuscular systems, this would represent
perturbations that the system might encounter, (b) Block diagram for an adaptive controller incorporating reinforcement

learning. In neuromuscular systems, the controller, plant and sensor will be the same as in (a). The critic will be the input from
all the interneurons, e.g. la, lb, Renshaw cells, etc. affecting the efficacy and excitability of the motor neuron (the controller),
which is represented by the reinforcement signals.

decide how and where the next increment in stepping number that scores the controller's behaviour; the
will be taken, choosing a path that will give the highest higher the number, the better the behaviour. If the
probability for future success. In this modern concept behaviour being scored is immediately preceding a
of reinforcement learning, randomness is often utilized subsequent unit of behaviour produced by the con
to create behavioural varieties, which are call explora troller based on the critic's score, there must be enough
tions. The consequential actions, however, are strongly variability in the controller's behaviour so that the critic
can to work.
guided by the evaluation of earlier experiences and evaluate many alternatives for this process
often the system will prefer an option that has produced A learning mechanism must then adjust the controller's

favourable results in the past; such a move is call behaviour so that it tends towards behaviours that are
As a result, favoured by the critics (Sutton & Barto 1998).
exploitation. reinforcement-learning algo
rithms are
selection processes, but there must be Applying these learning theories to the biological
in the so that system, it is apparent that it is necessary to have a
variability action-generation process
the consequences of alternative actions can be system where there is variability for unsupervised
to to occur.
compared select the best alternative. learning

By incorporating reinforcement learning into a

control system, such a system is able to use feedback
to evaluate the performance and improve subsequent 8. MERGING OF SENSORY AND ROBOTIC
movement by changing the controller itself (figure 56). CONTROL
This is analogous to sensory inputs changing the As de Leon et al have demonstrated, robotic devices can
efficacy of spinal circuitries that control locomotion as be used as effective training tools for SCI subjects (de
a result of training. In most artificial reinforcement Leon et al 2002a,b) and commercially available robotic
learning systems, the critic's output at any time is a orthotics, such as the Lokomat, are already available to

Phil Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 Spinal cord plasticity after injury L. L. Cai and others 1643

15
desired trajectory
10 boundary_

fr
2 0

"5
?
-10

-15
-20 -15 -10 -5 0 5 10 15 20
X trajectory (mm) 35 40 45 50 55 60
Figure 6. Schematic of a semi-active fixed-trajectory para
Figure 7. Soft robotic control schematics on how the semi
digm for step training, where the desired limb trajectory active control paradigm for step training is implemented. A
(blue) is bounded by both the inner and the outer boundaries moving window (red) bounds the desired trajectory (blue) of
(red). The actual trajectory (black) that the neural circuits the mouse limb during stepping. Within the window, the
might induce is allowed to vary within the boundary. robotic arm allows the mouse to vary its movement. However,
However, once the trajectory falls outside the boundary, the when the neural control desired trajectory falls outside the
robot will actively bring it back within the boundaries. The the robot will a convergent field
window, experience velocity
black line with periodic dots illustrates the potential positions that actively returns the mouse's limbs back within the
that the intrinsic neural control might choose to generate for window. This type of soft control is thought to approximate
any given bin time. The probability that the neural control as needed'
the 'assist approach used by experienced therapists
would move the limb to the exact position defined the blue
by (modified from Cai et al 2005).
line, representing a fixed trajectory, is highly unlikely. As a
result, theoretically, the neural control system is continuously

disrupted by the fixed trajectory paradigm. This fixed

does not allow the neural control arm
trajectory, therefore, defined by the robot control. Secondly, fixed
circuitry to respond to any of its intrinsic activation patterns,
trajectory training seems to eventually cause an
but rather forces the intrinsic circuitry to continuously
extensive level of habituation to the sensory information
respond to external perturbations. This strategy for control
generated, to the point that there is little or no response
would seem to unnecessarily disrupt the spinal circuitry and
in the minimize or even the intrinsic
to the invariant sensory input imposed by the robotic
process preclude
circuitry from interpreting relevant proprioceptive infor training device and, therefore, little or no motor output
mation to generate a solution is generated, i.e. a state of 'learned This
required (i.e. make choices) helplessness'.
and, thus, presumably prevent the circuitry from meaningful would result in a less robust, rather than an enhanced,

learning phenomena. motor response (Wirz et al 2005). Thirdly, fixed

trajectory training does not allow the damaged spinal
circuitry to learn to regain autonomous control, i.e.
facilitate the rehabilitative training of SCI and stroke formulate high probability, temporally important sen
patients, and have shown promising results (Hesse et al
sory motor synaptic events that generate stepping.
2003). One strategy for robotically training an SCI We have found a more effective method using
subject is to generate a fixed pattern that mimics the
robotics to enhance of locomotor control
recovery
kinematics and trajectory of a normal step (centre blue
largely by mimicking some of the basic principles used
bold line in figure 6). This idealized step could be
by physical therapists, such as 'assist as needed'
derived by generating amean kinematic and mechanical
training strategies. Training with robotic control
training trajectory based on a number of steps in an algorithms that provide an assist as needed approach
intact, normal animal (fixed trajectory training). In resulted in a greater in stepping than
improvement
other words, every step would be mechanically identical. with a fixed pattern. We refer to this principle
training
In such a situation, the spinal circuitry of a complete of robotic control as 'soft control', which is graphically
spinal subject would essentially have no control of the illustrated in figure 7. Conceptually, the robotic arms
step trajectory imposed by the robotic arms. This type of generate a force field that pushes the leg towards the
control for training subjects with a complete SCI, is greater than a
target trajectory when the deviation
theoretically, may be counterproductive. There may be
predetermined amount. The magnitude of the force
several fundamental reasons for this to be counter field is proportional to the distance away from the
productive. Firstly, a complete robotic control does not desired trajectory that the leg has moved. In essence,
accommodate the variability intrinsic to the neural the robot performs similarly to a therapist providing
circuits that normally generate stepping (mentioned manual assistance to a patient stepping on the treadmill
earlier). Consequently, at any given phase of the step with a weight-supporting device. With the 'soft control'
cycle, it is unlikely that the motor pattern, formed on the approach, the greater the deviation from the desired
basis of the intrinsic motor pattern combined with the trajectory, the more the assistance needed to correct the

sensory patterns associated with each phase of the step trajectory. When adult, complete spinal mice trained
cycle, would exactly match the idealized trajectory with this soft control paradigm were compared with

Phil. Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 1644 L. L. Cai and others Spinal cord plasticity after injury

be activated appropriately to produce a successful

step rather than simply by the presence of ana
tomical connectivity. Such functional flexibility in the
activation of the sensorimotor circuits for stepping,
in turn, would be responsible for the variability
inherent to gait patterns. This variability, however,
reflects a fundamental feature of the neural control
system that should be recognized and accommodated
in developing strategies designed to enhance motor
performance by motor training using robotic devices
after an SCI.

This work was supported, in part, by NIH grant NS16333.

48 50 52 54 56 58 60 62
X position (mm) REFERENCES
(b) Anderson, J. R., Michalski, R. S., Carbonell, J. G. & Mitchell,
T. M. 1983 Machine learning: an artificial intelligence
approach. Los
Altos, CA: M. Kaufmann.

Antri, M., Orsal, D. & Barthe, J. Y. 2002 Locomotor recovery

in the chronic spinal rat: effects of long-term treatment
with a 5-HT2 agonist. Eur. J. Neurosci. 16, 467-476.

(doi:10.1046/j.1460-9568.2002.02088.x)
Arshavsky, Y. I., Deliagina, T. G. & Orlovsky, G. N. 1997
Pattern generation. Curr. Opin. Neurobiol. 1, 781-789.
o
Oh (doi:10.1016/S0959-4388(97)80136-5)
Barbeau, H. & Rossignol, S. 1987 Recovery of locomotion
after chronic spinalization in the adult cat. Brain Res. 412,
84-95. (doi:10.1016/0006-8993(87)91442-9)
Barbeau, H., Fung, J., Leroux, A. & Ladouceur, M. 2002 A
review of the adaptability and recovery of locomotion after
44 46 48 50 52 54 56 58 60 62 64 66
spinal cord injury. Prog. Brain Res. 137, 9-25.
X positions (mm) Barto, A. G. 1994 Reinforcement learning control. Curr.

Opin. Neurobiol 4, 888-893. (doi:10.1016/0959-4388

Figure 8. (a) Trajectory plot of the ankle of an untrained adult
transected mouse without administration
(94)90138-4)
any drug attempt
Bianchi, L., Angelini, D., Orani, G. P. & Lacquaniti, F. 1998
ing to step on amoving treadmill for 10 s at a rate of 3 cm s ~x.
Kinematic coordination in human relation to
gait:
(6) Trajectory plot of the ankle of an adult transected mouse
mechanical energy cost. J. Neurophysiol. 79, 2155-2170.
successfully stepping on a moving treadmill for 10 s at a rate
Bregman, B. S., Diener, P. S., McAtee, M., Dai, H. N. &
of 3 cm s_1 after four weeks of step training with quipazine
James, C. 1997 Intervention strategies to enhance
(0.5 mg kg"l) daily for 10 min d_ l, 5 days per week. Arrows
anatomical and recovery of function after spinal
plasticity
in (6) are showing the direction of ankle movements. Note the
cord injury. Adv. Neurol. 72, 257-275.
more consistent trajectories in the trained versus untrained
Bregman, B. S., Coumans, J. V, Dai, H. N., Kuhn, P. L.,
mouse. The untrained mouse often failed to execute any
Lynskey, J., McAtee, M. & Sandhu, F. 2002 Transplants
successful plantar placing steps as shown in (6).
and neurotrophic factors increase regeneration and

a fixed trajectory recovery of function after spinal cord injury. Prog. Brain
spinal mice trained with training Res. 137, 257-273.
paradigm, the soft control trained mice stepped Burke, R. E. & Edgerton, V. R. 1975 Motor unit properties
significantly better (figure 8). The fact that the soft in movement. Exerc.
and selective involvement Sport Sei.
control paradigm was more effective than the fixed Rev. 3,31-81.
control paradigm is consistent with the many demon Cai, L. L., Fong, A. J., Otoshi, C. K, Liang, Y. Q., Cham,
strations that variability is a critical component that J. G., Zhong, H., Roy, R. R, Edgerton, V. R. & Burdick,
characterizes all neurally controlled movements. J. W. 2005 Effects of consistency vs. variability in

robotically controlled training of stepping in adult spinal

mice. Proc. Int. Conference Rehab. Robotics 9, 575-579.
9. CONCLUSION Cope, T. C. & Sokoloff, A. J. 1999 Orderly recruitment tested
In this paper, we have emphasized the high degree of across muscle boundaries. Prog. Brain Res. 123, 177-190.

plasticity of the functional connectivity within the Cote, M. P. & Gossard, J. P. 2004 Step training-dependent
sensorimotor infrastructure in to an plasticity in spinal cutaneous pathways. J. Neurosci. 24,
spinal response
11 317-11 327. (doi:10.1523/JNEUROSCI.1486-04.2004)
injury and/or step training. We have pointed out that
Cote, M. P., Menard, A. & Gossard, J. P. 2003 Spinal cats on
the neural processes involved in the generation of
the treadmill: changes in load pathways. J. Neurosci. 23,
standing and stepping are extremely flexible func
2789-2796.
tionally, and are unlikely to be due to a hardwired,
G. & M. 2004 of a basic
Courtine, Schieppati, Tuning
fixed neuronal architecture. Instead, there are many
coordination pattern constructs straight-ahead and curved
possible pathways and combinations of circuits that in humans. 1524-1535.
walking J. Neurophysiol. 91,
can generate movement. This view implies that
(doi: 10.1152/jn.00817.2003)
locomotor-related neural circuits are better defined Courtine, G., Roy, R. R., Hodgson, J., McKay, H., Raven,
as the probability of a given assembly of synapses to J., Zhong, H., Yang, H., Tuszynski, M. H. & Edgerton,

Phil Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 Spinal cord plasticity after injury L. L. Cai and others 1645

V. R. 2005a Kinematic and EMG determinants in Fong, A. J., Cai, L. L., Otoshi, C. K, Reinkensmeyer,
quadrupedal locomotion of a non-human primate D. J., Burdick, J. W., Roy, R. R. & Edgerton, V. R.

(Rhesus). J. Neurophysiol 93, 3127-3145. (doi:10. 2005 Spinal cord-transected mice learn to step in

1152/jn.01073.2004) response to quipazine treatment and robotic training.

Courtine, G., Roy, R. R., Raven, J., Hodgson, J., McKay, H., J. Neurosci. 25, 11738-11747. (doi:10.1523/JNEUR
Yang, H., Zhong, H., Tuszynski, M. H. & Edgerton, V. R. OSCI.1523-05.2005)
20056 Performance of locomotion and foot grasping Georgopoulos, A. P. & Grillner, S. 1989 Visuomotor

following a unilateral thoracic corticospinal tract lesion coordination in reaching and locomotion. Science 245,
in monkeys (Macaca mulatto). Brain 128, 2338-2358. 1209-1210.

(doi: 10.1093/brain/awh604) Gerasimenko, Y. P., Avelev, V. D., Nikitin, O. A. & Lavrov,

Cytowic, R. E. 2002 Synesthesia: a union of the senses. I. A. 2003 Initiation of locomotor activity in spinal cats by
Cambridge, MA: MIT Press, epidural stimulation of the spinal cord. Neurosci. Behav.
de laTorre, J. C. & Goldsmith, H. S. 1990 Collagen-omental Physiol 33, 247-254. (doi: 10.1023/A: 1022199214515)
graft in experimental spinal cord transection. Acta Neu Gerasimenko, Y P., Lavrov, I. A., Bogacheva, I. N.,
rochir (Wien) 102, 152-163. (doi: 10.1007/BFO 1405432) Shcherbakova, N. A., Kucher, V. I. & Musienko, P. E.
de Leon, R. D., Hodgson, J. A., Roy, R. R. & Edgerton, V. R. 2005 Formation of locomotor patterns in decerebrate cats
1998a Full weight-bearing hindlimb standing following in conditions of epidural stimulation of the spinal cord.
stand training in the adult spinal cat. J. Neurophysiol. 80, Neurosci. Behav. Physiol. 35, 291-298.
83-91. Grasso, R., Ivanenko, Y P., Zago, M., Molinari, M.,
de Leon, R. D., Hodgson, J. A., Roy, R. R. & Edgerton, V. R. Scivoletto, G., Castellano, V, Macellari, V. &
19986 Locomotor capacity attributable to step training Lacquaniti, F. 2004 Distributed plasticity of locomotor
versus spontaneous recovery after spinalization in adult pattern generators in spinal cord injured patients. Brain
cats.
J. Neurophysiol. 79, 1329-1340. 127, 1019-1034. (doi:10.1093/brain/awhll5)
de Leon, R. D., Tamaki, H., Hodgson, J. A., Roy, R. R. & Grillner, S. 2002 The spinal locomotor CPG: a target after
Edgerton, V. R. 1999 Hindlimb locomotor and postural spinal cord injury. Prog. Brain Res. 137, 97-108.
training modulates glycinergic inhibition in the spinal Grillner, S. 2003 The motor infrastructure: from ion
cord of the adult spinal cat. J. Neurophysiol. 82, channels to neuronal networks. Nat. Rev. Neurosci. 4,
359-369. 573-586. (doi:10.1038/nrnll37)
de Leon, R. D., Kubasak, M. D.,
Phelps, P. E., Timoszyk, Gurfinkel, V. S., Levik, Y S., Kazennikov, O. V. & Selionov,
W. K., Reinkensmeyer, D. J., Roy, R. R. & Edgerton, V. R. V. A. 1998 Locomotor-like movements evoked by leg
2002a Using robotics to teach the spinal cord to walk. muscle vibration in humans. Eur. J. Neurosci. 10,
Brain Res. Brain Res. Rev. 40, 267-273. (doi: 10.1016/ 1608-1612. (doi:10.1046/j.l460-9568.1998.00179.x)
S0165-0173(02)00209-6) Hase, T., Kawaguchi, S., Hayashi, H., Nishio, T., Mizoguchi,
de Leon, R. D., Reinkensmeyer, D. J., Timoszyk, W. K., A. & Nakamura, T. 2002 Spinal cord repair in neonatal

London, N. J., Roy, R. R. & Edgerton, V. R. 20026 rats: a correlation between axonal regeneration and
Use of robotics in assessing the adaptive capacity of functional recovery. Eur. J. Neurosci. 15, 969-974.
the rat lumbar spinal cord. Prog. Brain Res. 137, (doi:10.1046/j.l460-9568.2002.01932.x)
141-149. Henneman, E. & Mendell, L. M. 1981 Functional organiz
Dimitrijevic, M. R, Gerasimenko, Y. & Pinter, M. M. 1998 ation of motoneuron pool and itsinputs. In Handbook of
Evidence for a spinal central pattern generator in humans. physiology. The Nervous System. Motor Control, section 1,
Ann. NYAcad. Sei. 860, 360-376. (doirlO.l 11 l/j.1749 vol. II, pt. 1, ch. 11, pp. 423-507. Bethesda, MD:

6632.1998.tb09062.x) American Physiological Society.

Drew, T., Jiang, W. & Widajewicz, W. 2002 Contributions Hesse, S., Schmidt, H., Werner, C. & Bardeleben, A.
of the motor cortex to the control of the hindlimbs 2003 Upper and lower extremity robotic devices for
during locomotion in the cat. Brain Res. Brain Res. Rev. rehabilitation and for studying motor control. Curr.
40, 178-191. (doi:10.1016/S0165-0173(02)00200-X) Opin. Neurol. 16, 705-710. (doi:10.1097/00019052
Edgerton, V. R, Roy, R. R, Hodgson, J. A., Prober, R. J., 200312000-00010)
de Guzman, C. P. & de Leon, R. 1992 Potential of Ichiyama, R. M., Gerasimenko, Y P., Zhong, H., Roy, R. R.
adult mammalian lumbosacral spinal cord to execute & Edgerton, V. R. 2005 Hindlimb stepping movements in
and acquire improved locomotion in the absence complete spinal rats induced by epidural spinal cord
of supraspinal input. J. Neurotrauma 9(Suppl. 1), stimulation. Neurosci. Lett. 383, 339-344. (doi: 10.1016/
S119-S128. j.neulet.2005.04.049)
Edgerton, V. R. et al. 2001a Retraining the injured spinal Kuiken, T. A., Dumanian, G. A., Lipschutz, R. D., Miller,
cord. J. Physiol 533, 15-22. (doi:10.1111/j. 1469-7793. L. A. & Stubblefield, K. A. 2004 The use of targeted

2001.0015b.x) muscle reinnervation for improved myoelectric prosthesis

Edgerton, V. R, Roy, R. R. & De Leon, R. 20016 Neural control in a bilateral shoulder disarticulation amputee.
Darwinism in the mammalian spinal cord. Spinal cord Prosthetist Orthotist Int. 28, 245-253.
plasticity: alterations in reflex function. Boston, UK: Kluwer Kuypers, H. G. 1978 The motor system and the capacity to
Academic Publishers. execute highly fractionated distal extremity movements.

Edgerton, V. R., Tillakaratne, N. J., Bigbee, A. J., de Leon, Electroencephalogr. Clin. Neurophysiol. 34, 429-431.
R. D. & Roy, R. R 2004 Plasticity of the spinal neural Lacroix, S., Havton, L. A., McKay, H., Yang, H., Brant, A.,
circuitry after injury. Annu. Rev. Neurosci. 27, 145-167. Roberts, J. & Tuszynski, M. H. 2004 Bilateral corticosp
(doi: 10.1146/annurev.neuro.27.070203.144308) inal projections arise from each motor cortex in the

Fetz, E. E., Perlmutter, S. I. & Prut, Y. 2000 Functions of macaque monkey: a quantitative study. J. Comp. Neurol.
mammalian spinal interneurons during movement. Curr. 473, 147-161. (doi:10.1002/cne.20051)
Opin. Neurobiol 10, 699-707. (doi:10.1016/S0959-4388 Lafreniere-Roula, M. & McCrea, D. A. 2005 Deletions of

(00)00160-4) rhythmic motoneuron activity during fictive locomotion

Fong, A. J. et al 2003 Effects ofquipazine and robotic training on and scratch provide clues to the organization of the

spinal mice. Washington, DC: Society of Neuroscience mammalian central pattern generator. J. Neurophysiol.
Abstract Program No. 498.20. 94, 1120-1132.

Phil. Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions
 1646 L. L. Cai and others Spinal cord plasticity after injury

Lawrence, D. G. & Kuypers, H. G. 1968 The functional Scott, J. G. & Mendell, L. M. 1976 Individual EPSPs
of the motor system in the monkey. I. The produced by single triceps surae la afferent fibers in
organization
effects of bilateral pyramidal lesions. Brain 91, 1-14. homonymous and heteronymous motoneurons.

Lee, Y S., Lin, C. Y, Robertson, R. T., Hsiao, I. & Lin, V. W. 39, 679-692.
J. Neurophysiol.
2004 Motor recovery and anatomical evidence of axonal M. F. V. & Orlovskii, G. N. 1966 Control of
Shik, L., Severin,
in spinal adult rats. J. Neuropathol
regrowth cord-repaired walking and running by means of electric stimulation of
233-245.
Exp. Neurol 63, the midbrain. Biofizika 11, 659-666.
Lemon, R. N., Kirkwood, P. A., Maier, M. A., Nakajima, K. A. & P. 1976 Muscle control
Sjostrom, Zangger, spindle
& Nathan, P. 2004 Direct and indirect pathways for movements the deaf
during locomotor generated by
corticospinal control of upper limb motoneurons in the
ferented spinal cord. Acta Physiol. Scand. 97, 281-291.
primate. Prog. Brain Res. 143, 263-279. - an
Sutton, R. S. & Barto, A. G. 1998 Reinforcement learning
Lovely, R. G., Gregor, R. J., Roy, R. R. & Edgerton, V. R.
introduction. Cambridge, MA: MIT Press.
1986 Effects of training on the recovery of full-weight
Thompson, F J., Parmer, R. & Reier, P. J. 1998 Alteration in
bearing in the adult spinal cat. Exp. Neurol. 92,
stepping rate modulation of reflexes to lumbar motoneurons after
421-435. (doi:10.1016/0014-4886(86)90094-4) midthoracic spinal cord injury in the rat. I. Contusion
Lovely, R. G., Gregor, R. J., Roy, R. R. & Edgerton, V. R.
J. Neurotrauma
injury. 15, 495-508.
1990 Weight-bearing hindlimb stepping in treadmill
Tillakaratne, N. J., Mouria, M., Ziv, N. B., Roy, R. R.,
exercised adult spinal cats. Brain Res. 514, 206-218.
Edgerton, V. R. & Tobin, A. J. 2000 Increased
(doi:10.1016/0006-8993(90)91417-F)
of expression of glutamate decarboxylase (GAD (67)) in
MacKay-Lyons, M. 2002 Central pattern generation
feline lumbar spinal cord after complete thoracic spinal
locomotion: a review of the evidence. Phys. Ther. 82, 69-83.
A. G. 1972 Classification cord transection. J. Neurosci. Res. 60, 219-230. (doi: 10.
Orlovsky, G. N. & Feldman, of
lumboscaral neurons according to their discharge patterns 1002/(SICI)1097-4547(20000415)60:2<219::AID-JNR
during evoked locomotion. Neurophysiology. New York, NY: 11>3.0.CO;2-F)
Plenum. Tillakaratne, N. J., de Leon, R. D., Hoang, T. X., Roy, R. R.,

M., Feraboli-Lohnherr, D., Edgerton, V. R. & Tobin, A. J. 2002 Use-dependent

Orsal, D., Barthe, J. Y, Antri,
Yakovleff, A., Gim?nez y Ribotta, M., Privat, A., modulation of inhibitory capacity in the feline lumbar
Provencher, J. & Rossignol, S. 2002 Locomotor recovery spinal cord. J. Neurosci. 22, 3130-3143.
in chronic rat: pharmacological treat M., Danilov, Y & Bach, Y R. P. 2003 Closing an open
spinal long-term Tyler,
ment or transplantation of embryonic neurons? Prog. Brain loop control system: vestibular substitution through the
Res. 137, 213-230. tongue. J. Integr. Neurosci. 2, 159-164. (doi: 10.1142/

Pearson, K. G. 2004 Generating the walking gait: role of

S0219635203000263)
feedback. Prog. Brain Res. 143, 123-129. X. 2004a
sensory Valero-Cabre, A., Fores, J. & Navarro, Reorgan
Petruska, J. C, Ichiyama, R. M., Crown, E., Kansey, K., ization of reflex mediated different afferent
responses by
V. R. & Mendell, L. M. 2004 Segmental and
Edgerton, sensory fibers after spinal cord transection. J. Neurophysiol
central to motoneurons change following spinal cord
inputs 91, 2838-2848. (doi:10.1152/jn.01177.2003)
transection and step training in rats. Washington, DC:
Valero-Cabre, A., Tsironis, K, Skouras, E., Navarro, X. &
Society of Neuroscience Abstract Program No. 418.10.
Neiss, W. F. 20046 Peripheral and spinal motor reorgan
Raineteau, O. & Schwab, M. E. 2001 Plasticity of motor
ization after nerve injury and repair. J. Neurotrauma 21,
systems after incomplete spinal cord injury. Nat. Rev.
95-108. (doi:10.1089/089771504772695986)
Neurosci. 2, 263-273. (doi: 10.1038/35067570) A. & Cagol, E. 1995
Wernig, A., M?ller, S., Nanassy,
Robinson, G. A. & Goldberger, M. E. 1986a The
Laufband therapy based on 'rules of spinal locomotion'
development and recovery of motor function in spinal
is effective in spinal cord injured persons. Eur. J.
cats. I. The infant lesion effect. Exp. Brain Res. 62,
Neurosci. 1, 823-829. (doi: 10.1111/j. 1460-9568.1995.
373-386.

Robinson, G. A. & Goldberger, M. E. 19866 The develop

tb00686.x)
cats. Wirz, M., Hornby, R.5 Rupp, R. & Dietz, V. 2005 Locomotor
ment and recovery of motor function in spinal II.
Res. training with a driven gait orthosis in incomplete spinal
Pharmacological enhancement of recovery. Exp. Brain

62, 387-400.
cord injury. Gait Posture 21, S74. (doi:10.1016/S0966
Rossignol, S. & Barbeau, H. 1993
Pharmacology of 6362(05)80242-5)
locomotion: an account of studies in spinal cats and spinal Zhou, K, Doyle, J. C. & Glover, K. 1996 Robust and optimal
Soc. 16, 190-196. control. Saddle Prentice Hall.
cord injured subjects. J. Am. Paraplegia Upper River, NJ:

Phil. Trans. R. Soc. B (2006)

This content downloaded from 128.235.251.160 on Sat, 17 Jan 2015 03:16:35 AM

All use subject to JSTOR Terms and Conditions