Brain (2002), 125, 1150±1161
Interlimb re¯exes and synaptic plasticity become
evident months after human spinal cord injury
Blair Calancie, Maria R. Molano and James G. Broton
The Miami Project to Cure Paralysis and Department of
Neurological Surgery, University of Miami School of
Medicine, Miami, FL 33136, USA
Summary
Persons with long-standing injury to the cervical spinal
cord resulting in complete or partial paralysis typically
develop a wide spectrum of involuntary movements in
muscles receiving innervation caudal to the level of
injury. We have previously shown that these movements
include brief and discrete contraction of muscles in the
hand and forearm in response to innocuous sensory
stimulation to the feet and legs, but we have been
unable to replicate these interlimb re¯exes in ablebodied subjects. Properties of these muscle responses
indicate that the synaptic contacts between ascending
sensory ®bres and motor neurones of the cervical
enlargement are more ef®cacious than normal. If these
connections are present at all times, and require the
more rostrally-placed spinal cord injury to allow their
emergence, one might expect their appearance relatively
soon following injury, as has been shown for studies of
`latent' synapses. Conversely, delayed appearance of
these interlimb re¯exes would suggest either the development of new synaptic connections or a profound
strengthening of existing circuits in the cervical spinal
cord due to a combination of afferent target loss and
motor neurone denervation from motor tracts originat-
Correspondence to: Blair Calancie, PhD, Department of
Neurosurgery, SUNY's Upstate Medical University, 750
East Adams Street, IHP Room 1213, Syracuse, NY 13210,
USA
E-mail: [email protected]
ing rostral to the injury site. In this study, we used
repeated examinations of persons with acute injury to
the cervical spinal cord to examine the time post-injury
at which interlimb re¯exes are ®rst seen. Using tibial
nerve stimulation at the knee as a screening test, a total
of 24 subjects were found to develop interlimb re¯exes
following spinal cord injury. Latencies between stimulation and EMG were as brief as 32 ms for muscles of
the forearm and 44 ms for muscles in the hand. These
minimal delays all but rule out a supraspinal route for
these interlimb re¯exes. Interlimb re¯exes ®rst became
evident no sooner than ~6 months following injury, and
in some individuals were not seen until well over 1 year
post-injury. Enhanced lower limb segmental excitability
had emerged in nearly all of these subjects weeks or
months prior to the ®rst appearance of interlimb
re¯exes, arguing against a manifestation of traditional
post-traumatic spasticity as a basis for this activity.
This prolonged delay between time of injury and emergence of interlimb re¯ex activity lends support to the
hypothesis that this activity represents an example of
plasticityÐand perhaps `regenerative sprouting'Ðin
the human spinal cord following traumatic injury.
Keywords: spinal cord injury; plasticity; regenerative sprouting; re¯ex; human; cervical
Abbreviations: ADM = hypothenar group of the hand; APB = thenar group of the hand; ASIA = American Spinal Injury
Association; ECR = wrist extensors; FCR = wrist ¯exors; ILR = interlimb re¯exes; Psoas = hip ¯exors; SCI = spinal cord
injury
Introduction
There are numerous examples of axonal regeneration and
synaptic reorganization of neurones (collectively referred to
as `plasticity') following traumatic lesions to the mammalian
spinal cord (for a review, see Guth, 1974; Steward, 1989;
Goldberger et al., 1993; Schwab and Bartholdi, 1996;
Mendell et al., 2001; Siddall and Loeser, 2001; Wolpaw
and Tennissen, 2001). In the absence of speci®c intervenã Guarantors of Brain 2002
tions, such plasticity may contribute to the development of
abnormal movement states (McCouch et al., 1958; Nelson
and Mendell, 1979; Hiersemenzel et al., 2000) and/or sensory
disturbances (Christensen and Hulsebosch, 1997; Romero
et al., 2000). Alternatively, certain post-injury alterations of
behaviour may re¯ect the collective action of synaptic
connections that were present at the time of injury, but
Interlimb re¯exes after spinal cord injury
were functionally inactive (or `silent'). These latent connections have been reported to become `unmasked' within
minutes to hours of a CNS lesion (Goshgarian and Guth,
1977; Nelson et al., 1979; Wall, 1988; Goshgarian et al.,
1989; but compare with Brown et al., 1984). Thus the
timecourse with which certain spinal cord input/output
properties emerge following CNS trauma may provide clues
as to the mechanism(s) underlying those behaviours.
Many of the same mechanisms of plasticity reported from
animal models of spinal cord injury (SCI) can likely be
demonstrated in human subjects, but quantitative histologic
data supporting the possibility of sprouting or synaptogenesis
are indirect (Krassioukov et al., 1999). Behavioural studies
abound though and raise the possibility for both spontaneous
(Calancie et al., 1994) and task-speci®c (Bach and Rita, 1981;
Wernig et al., 1995; Harkema et al., 1997; Barbeau et al.,
1999) plasticity in the human spinal cord caudal to an injury.
Previous reports from this laboratory (Calancie, 1991;
Calancie et al., 1996) have described novel `interlimb
re¯exes' (ILR) in persons who have sustained SCI >1 year
prior to study (i.e. in the chronic phase). These involuntary
movements are characterized by short-latency (i.e. 40±50 ms)
contractions of hand and forearm muscles following a wide
variety of innocuous sensory stimuli delivered to the lower
limb or limbs (including skin stroking, hair pull, tendon taps
and electrical stimulation of peripheral nerves). We have
suggested that such interlimb re¯exes may re¯ect the
consequences of novel synaptic connections formed between
ascending ®rst- and second-order afferent ®bres and motor
neurones of the cervical enlargement partially denervated due
to a more rostrally-placed lesion to the spinal cord (Calancie
et al., 1996). If correct, one would expect to see a signi®cant
delay between the time of injury and the time at which such
ILRs become evident.
In this paper, we report ®ndings from a group of subjects
who ultimately developed interlimb re¯exes after sustaining
traumatic spinal cord injury. We conducted repeated measures on these subjects over a period of many months to
determine the time after injury when these re¯exes became
evident. Our data are consistent with the hypothesis that these
interlimb re¯exes represent the establishment of new synaptic
connections between nerve populations that do not normally
interact. We suggest that ILR emergence serves as an
example of CNS plasticity (or `regenerative sprouting';
Steward, 1989) in the adult human nervous system following
traumatic injury. This conclusion is included within a much
broader examination of plasticity after human SCI that was
presented previously (Calancie et al., 2000).
Methods
Subjects
Experiments were performed on persons with traumatic
injury to the cervical spine resulting in neurologic de®cit (i.e.
spinal cord injury). In the majority of cases, the initial
1151
examination of a given subject took place within the ®rst
week after injury, with follow-up studies continuing over the
next weeks and months post-injury. All subjects gave their
informed consent to participate in this protocol, which was
approved by the University of Miami's Institutional Review
Board.
Procedures
Self-adhesive surface EMG electrodes (S'Offset; Graphic
Controls Corp., Buffalo NY, USA) were positioned over the
biceps brachii, triceps brachii, wrist extensors (ECR), wrist
¯exors (FCR), thenar group of the hand (APB), hypothenar
group of the hand (ADM), hip ¯exors (Psoas), quadriceps,
hamstring, tibialis anterior (TA), soleus, and foot intrinsics of
the subject's left side. Examinations were made of voluntary
individual muscle contractions, tendon re¯exes and central
motor conduction in response to transcranial magnetic
stimulation, and interlimb re¯exes through surface-applied
electrical stimulation of the tibial nerve at the popliteal fossa.
Tibial stimulation was accomplished using either a Grass
S88 stimulator (initial four subjects; Grass Instrument Co,
Quincy, MA, USA) or a Digitimer D185 stimulator (remaining 20 subjects; Digitimer Inc, Welwyn Garden City, UK).
Stimuli were delivered through pairs of self-adhesive surface
electrodes (Cleartrace; ConMed Corp., Utica, NY, USA)
positioned over the tibial nerve (cathode) and a site medial
and distal to this site (anode). Single pulses of >100 V were
used to de®ne the optimal stimulus site, using soleus direct
muscle response (M-wave) and plantar-going ankle movement as the criteria to guide stimulation. In most cases,
pressure was applied to the cathode while stimuli were
delivered, pushing the electrode closer to the underlying tibial
nerve in order to minimize subject discomfort (in those
subjects who could feel the stimulus) while still eliciting a
strong plantar-¯exion. The stimulus intensity delivered via
the D185 stimulator [based on readings from the stimulator's
liquid crystal display (LCD)] was not less than 125 mA in any
subject tested, while trials using the Grass stimulator
routinely used pulses of 150 V (the maximum capable with
this device). The duration of individual stimuli within a
3-pulse train was 50 ms and 1000 ms for the D185 and S88
stimulators, respectively. (Note that the D185 pulse width
cannot be adjusted from this 50 ms value, but that its
maximum stimulus intensity far exceeds that of the S88,
enabling the D185 output to produce strong plantar-going
twitches when desired.)
Once an acceptable stimulation site was established, a
series of 3-pulse stimulus trains was delivered, each pulse in
the train separated from the next by 2 ms (i.e. three pulses at
2 ms each; `3 @ 2'). A 4-pulse train was used on occasion.
Pulse trains were separated from one another by a minimum
of 1 s. In most cases, the rate of pulse train stimulation was
approximately 0.2 Hz, and was controlled manually (i.e. a
deliberate button push was needed to trigger a stimulus). No
fewer than 10 stimulus trains were delivered, and the EMG
1152
B. Calancie et al.
responses from the six upper limb muscles were monitored
via computer (Toshiba T6400 and RC Electronics
Computerscope; RC Electronics Inc, Goleta, CA, USA) and
stored on digital tape [Vetter 4000a (defunct) or MicroData
Instruments DT1600; MicroData Instrument Inc, South
Plain®eld, NJ, USA]. After collecting left-side responses to
left-side stimulation, stimuli were applied to the right-side
tibial nerve while still recording from left-side muscles (and
the right-side soleus to con®rm response properties).
Additional evaluations were carried out, all electrodes were
shifted to the comparable right-side muscles, and the
measures repeated.
Data were analysed off-line (Cambridge Electronic Design
1401; Cambridge Electronic Design, Cambridge, UK) from
the taped records. EMG from each of the upper-limb muscles
being studied was offset to zero, recti®ed and averaged for a
200 ms post-stimulus time period. Minimum response
latencies were determined relative to the onset of the ®rst
stimulus in the pulse train. Responses were considered
positive if: (i) the response at threshold stimulus intensity
was lost if a 2-pulse train or single stimulus was applied; and
(ii) the response latency was <110 ms (more prolonged
latencies can be due to input from supraspinal elements, a
demonstration of which follows). In some cases of wellde®ned single motor unit (SMU) discharge, peri-stimulus
time histograms were produced to summarize that motor
unit's discharge properties across all stimuli. Template
matching was used to de®ne the motor unit's discharge
latency for each stimulus, the results of which were grouped
into 1 ms bins.
Muscle interference patterns were graded on a 6-point
scale, for which `0' represents a total absence of volitional
recruitment and `5' represents normal recruitment (this EMG
scale is fully described elsewhere by Alexeeva et al., 1997;
Calancie et al., 2001). Subjects who couldÐon commandÐ
recruit and silence EMG in at least one lower limb muscle
during attempted voluntary contractions and relaxations of
that muscle were termed `motor-incomplete' for the purposes
of this study.
Results
A total of 24 subjects in our series, all male, were found to
develop interlimb re¯exes after acute spinal cord injury.
Table 1 shows the cause of injury and the most rostral level of
fracture for each subject, along withÐat the time ILR activity
was ®rst notedÐeach subject's age, the most rostrallyinnervated upper limb muscle from which interlimb re¯exes
were seen, and the subject's clinical status as de®ned by
American Spinal Injury Association (ASIA) nomenclature
(Ditunno et al., 1994). Nineteen subjects were ®rst examined
within 10 days of their traumatic injury. Initial examinations
on subject numbers 1, 11, 12, 22 and 23 were conducted 21,
25, 18, 30 and 42 days after their respective injury dates.
These delays were due to transfer from another hospital (n = 4)
or an inability to consent due to sedation (n = 1). We
Table 1 Subject information
Subject number
Age (years)
Cause of injury
Rostral-most fracture
Rostral-most positive muscle
ASIA category
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
17
43
22
51
54
17
33
36
26
54
22
24
57
32
40
32
68
30
47
44
22
38
20
23
Diving
Diving
Gunshot
Blunt trauma
MVA
Sport
Fall
MVA
Fall
Diving
Gunshot
Diving
Sport
MVA
MVA
Fall
Blunt
MVA
MVA
PHBC
Fall
Diving
MVA
MVA*
C4
C5
C5
C3
C7
C4
C5
C4
C5
C5
C5
C5
C5
C4
C2
C5
C5
C5
C5
C4
C5
C2
C6
C4
ECR
ADM
FCR
Biceps brachii
APB
ADM
ECR
ADM
ADM
FCR
ECR
ECR
APB
APB
APB
APB
APB
APB
ADM
ECR
APB
ADM
ECR
FCR
A
C
A
C
C
A
A
C
A
A
A
A
A
A
A
D
C
D
C
C
A
C
C
C
Age, rostral-most positive muscle responding to stimulation and ASIA status were all based on the date when interlimb re¯exes were ®rst
seen following SCI. MVA = motor vehicle accident; MVA* = motorcycle accident; PHBC = pedestrian hit by car.
Interlimb re¯exes after spinal cord injury
1153
Fig. 2 From Subject 4. Superimposition of right-side ADM single
motor unit responses to left-side tibial nerve stimulation (upper
trace) and the averaged EMG following recti®cation (lower trace).
(A) Stimulation with a 3-pulse train (n = 12 trials). (B) Stimulation
with a 4-pulse train (n = 15 trials). Stimulus intensity was 200 mA
for both. Post-stimulus onset time of the EMG response is shown
on each averaged record. Vertical calibration = 50 mV.
Fig. 1 From Subject 4. Superimposition of 11 left-side EMG
records following 3-pulse stimulation (2 ms interstimulus interval)
to the right-side tibial nerve at average rates of 0.3 Hz (A; n = 30
trials) and 0.8 Hz (B; n = 33 trials). Direct muscle response
(M-wave) in the right-side soleus muscle can be seen arising out
of the stimulus artefacts in both A and B, con®rming adequate
stimulation. Single motor units were recruited in the biceps (one
unit only) and ADM (possibly two units; see text) muscles. The
amplitude of the biceps motor unit potential grew with successive
discharges following each stimulus, such that potentials late in
each trace are larger (peak-to-peak) compared with the ®rst
discharge following each stimulus train. This is readily evident
when observing single trials, but less so when presented as
overlapping traces. Post-stimulus time histograms representing
discharge of these motor units are shown at the bottom of each
panel. Stimulus intensity was 200 mA in all cases. Vertical
calibrations as follows: channels 1±5: 100 mV; 6: 50 mV; 7±10:
1 mV; 11±12: 10 mV. AbH = foot intrinsics; Biceps = biceps
brachii; Hams = hamstring; Quads = quadriceps.
attempted to perform follow-up examinations at least ®ve
times during the ®rst year after injury, but this was not always
possible (due to subject discharge, transfer to another facility,
moving out of town, incarceration, etc.). As few as three and
as many as 10 examinations for ILR activity were performed
on each subject in our population before such responses were
®rst seen (mean 6 standard deviation: 6.0 6 1.8).
Original and processed data records, which re¯ect several
elements of interlimb re¯exes consistent with our experience,
are illustrated in Figs 1 and 2, both taken from Subject 4.
Fig. 1 shows EMG (superimposed) from 11 left-side muscles
(and right-side soleus) in response to right-side tibial nerve
electrical stimulation with a 3-pulse train, at a slow (0.3 Hz;
Fig. 1A) and a somewhat faster (0.8 Hz; Fig. 1B) average
presentation rate. Each stimulus train resulted in short-latency
single motor unit recruitment in each of the (contralateral)
biceps and (contralateral) ADM muscles. Note that what
appear to be EMG responses in the triceps muscle were
probably caused by discharge in the biceps motor unit, due to
volume conduction (see Calancie et al., 2001). The poststimulus time histograms of the biceps and ADM motor units
are shown at the bottom of Fig. 1. Contraction of the biceps
tendon of insertion and movement of the intrinsic muscles
adjacent to the ®fth metacarpal was clearly visible with each
stimulus delivery, but no other left- or right-side upper limb
muscle contraction was evident in this subject in response to
right-side tibial nerve stimulation.
Of the biceps brachii and ADM motor unit records
illustrated in Fig. 1A, the ADM motor unit had the shorter
response latency, despite the longer conduction distances
involved. The response latency of the biceps motor unit
showed considerably more variability from trial to trial,
particularly for the slow stimulus rate. A modest increase in
the average rate of stimulation from 0.3 trains/s (Fig. 1A) to
0.8 trains/s (Fig. 1B) led to a more consistent response latency
of this biceps brachii motor unit, at a considerably shorter
latency (as early as 55 ms at the higher stimulus rate)
compared with its earliest latency of 60 ms at a slower rate of
stimulation. Similar ®ndings have been reported previously in
persons with chronic SCI (Calancie, 1991). This effect of
increasing stimulus rate was not readily apparent in the initial
response histogram for the ADM motor unit, but discharge at
1154
B. Calancie et al.
more prolonged latencies (~90 ms) following the stimulus
was more tightly grouped for the faster rate of stimulation in
this muscle. Note that close inspection of individual stimulusevoked responses for this ADM motor unit could not rule out
the possibility that a second motor unit of nearly identical
waveform shape contributed to the later periods (>80 ms) of
discharge in the histograms shown in Fig. 1A and B.
Figure 2 illustrates the behaviour of a single motor unit
from the right-side ADM muscle of Subject 4 in response to a
3 @ 2 (Fig. 2A) or 4 @ 2 (Fig. 2B) stimulus pattern to the left
(i.e. contralateral) tibial nerve, using a stimulus intensity of
200 mA. Because the rate of stimulation can affect a motor
unit's response probability and latency (as shown in Fig. 1),
we restricted the trials giving rise to Fig. 2 to stimulus train
intervals of 2.76±3.29 s. Using this approach, the mean
stimulus train delivery rate was 0.32 Hz for the 3-pulse train
(n = 12), and 0.34 Hz for the 4-pulse train (n = 15). Despite
these almost identical rates of train delivery, the response
latency onset for the ADM motor unit was clearly earlier for
the 4-pulse train (55.7 ms) compared with the 3-pulse train
(60.1 ms).
To compare their minimum reaction latency to latencies of
the interlimb re¯exes reported here, we asked some of our
subjects with partial sensation in their lower limbs to react
with a forceful voluntary extension of their wrist as soon as
they felt the electrical stimulus at the back of the knee. For
these trials, the audio for the EMG records was turned off and
subjects were asked to close their eyes in order to eliminate
auditory and visual cues of the stimulation. Figure 3 illustrates an example of this minimal reaction time in Subject 8
showing averaged, recti®ed EMG from multiple right-side
muscles in response to a 3 @ 2 stimulus train of 180 V
(216 mA) delivered to the left tibial nerve. In Figure 3A, the
subject was instructed to remain as relaxed as possible
throughout delivery of the 25 stimulus trains used. In Fig. 3B,
the subject was asked to extend his wrist (activating primarily
ECR) in response to each of the ®ve stimulus trains. The
minimum onset latencies of the interlimb re¯ex (~53 ms) and
volitional components (~115 ms) of the EMG responses are
indicated with arrows. For all subjects tested in this manner,
none could initiate clearly de®ned wrist extensor EMG at
latencies of <105 ms in response to tibial nerve stimulation;
most showed minimum onset latencies of 115±130 ms, well
beyond the latencies of the ILR activity we typically
observed.
A chief goal of the present study was to investigate the time
at which ILR activity became evident relative to each
subject's time of injury. Figure 4 summarizes our ®ndings,
showing for each subject the various examination times after
injury (in years) at which ILR activity was looked for and not
seen (interlimb re¯exes absent; open circles) and eventually
looked for and seen for the ®rst time (interlimb re¯exes
present; ®lled circles). Results have been sorted from shortest
to longest post-injury period when ILR activity was ®rst
observed. The subject numbers correspond to those used in
Table 1. Persons with a motor-incomplete injury (ASIA
category C or D) at the time at which ILR activity was ®rst
seen are indicated with asterisks.
The earliest time after which ILR activity could be evoked
by tibial nerve stimulation in this population was 5 months,
3 weeks post-injury (Subjects 1 and 2). For all subjects, the
average duration between date of injury and date of the ®rst
positive ILR observation was 447 6 221 days. This number
must be viewed with caution, however, since we were unable
to carry out follow-up evaluations at the same time postinjury across all subjects. For the same reason, we cannot be
nearly as certain as to how long post-injury ILR activity was
not seen, due again to the widespread variability in the timing
of follow-up examinations in this population. In most cases,
subjects were discharged from our hospital by 2 months postinjury, and for many of them securing transportation back to
our centre for follow-up studies on a regular basis was
problematic. Despite these dif®culties in the timing of followups, ILR activity was looked for, and found to be absent, as
late as 11 months post-injury in seven of the 24 subjects
included in this study. The longest delay between injury and
con®rmed absence of ILR activity was 1.6 years (Subject 23).
Did the time post-injury at which ILR activity became
evident correspond with the timing of the development of
hyperre¯exia and/or spasticity in these subjects? The answer
is no; in most subjects the emergence of ILR activity occurred
months after the recovery of spinal re¯exes, based upon the
following observations. First, all subjects who were grouped
in the motor-incomplete cohort had visible, brisk responses to
taps at the Achilles and/or patellar tendons from the very ®rst
examination onwards. Despite maintenance of spinal re¯ex
excitability early after their spinal injury, ILR activity was not
evident at this time in this cohort. Secondly, persons in both
groups (motor-complete and motor-incomplete SCI) showed
large increases in mean response amplitudes to taps at the
Achilles and patellar tendons before the emergence of ILR
activity. These data are summarized in Table 2, which shows
the average maximum peak-to-peak amplitudes of EMG
responses to Achilles and patellar tendon taps at the initial
evaluation, and at the most recent evaluation prior to that in
which ILR activity was ®rst seen. Even before the time when
ILR activity was ®rst observed, most subjects in both groups
showed well-de®ned responses to taps at both the patellar and
Achilles tendons. Also, 18 of the 24 subjects studied had
already been started on anti-spasmodic medication (typically
baclofen) at this time (i.e. prior to emergence of interlimb
re¯exes), suggesting that spasticity had already been noted by
their treating physicians.
Table 3 summarizes the distribution of muscles from which
ILR activity was observed following tibial nerve stimulation
in our subject population, and whether or not the site of
stimulation was on the same side as the site of evoked muscle
response or contralateral to the recruited muscle. Two trends
are evident from Table 3. First, responses tend to be
contralateral to the side stimulated [35 out of 49 instances
(71%)]. Secondly, ILR activity was far more commonly seen
in the intrinsic muscles of the hand, particularly the ADM
Interlimb re¯exes after spinal cord injury
Fig. 3 Averaged, recti®ed EMG records from Subject 8, who was
asked to: (A) remain relaxed in response to a series (n = 25) of
stimulus trains delivered to the contralateral tibial nerve; or (B)
forcefully extend his wrist as soon as he perceived the onset of
each of a series (n = 5) of stimulus trains. The vertical calibration
is 0.2 mV for traces from the biceps brachii, APB and ADM
muscles in both (A) and (B), and 2 mV for the remaining muscles
studied.
1155
muscle, than more proximal muscles in this subject population. In fact, only one case of response in the biceps brachii
muscle to tibial nerve stimulation was noted in this series.
The evoked response latencies (onset) to tibial nerve
stimulation are shown for the different upper limb muscles in
Table 4. In addition to the average (and standard deviation)
latency by muscle, the earliest and most prolonged latencies
seen in a given subject are also shown. Two conclusions can
be drawn from these data. First, the averaged response latency
increases with conduction distance from the cervical spinal
cord (i.e. distal hand muscle latencies are, on average, more
prolonged than those for muscles in the forearm). Secondly,
there is a considerable range of minimum response latencies
in a given muscle (excepting biceps brachii, of course, in
which only one positive response to tibial nerve stimulation
was seen in the 24 subjects examined). Given this range of
latencies, it is clear that the situation depicted in Fig. 1, in
which the latency to response in the ADM unit shown was
less than that in the biceps motor unit, is not totally
unexpected.
Based on a 6-point EMG recruitment scale (Calancie et al.,
2001), upper limb muscles in which ILR responses were seen
tended to be weaker (mean recruitment score = 1.9 6 2.3)
than those in which ILR activity was not seen (2.7 6 3.9); this
difference was statistically signi®cant. There were no cases in
which a muscle with normal EMG interference pattern (i.e. a
score of `5') demonstrated ILR activity. Furthermore, there
were no instances in which a single motor unit under
volitional control (i.e. it could be recruited and silenced at
Fig. 4 Summary of time from injury until the ®rst observation of ILR activity (positive response: ®lled
circles) and the latest time after injury when ILR activity was looked for, but not seen (negative response:
open circles) in the sample population. Subject number is consistent with that used in Table 1. Filled
circles with an asterisk represent motor-incomplete subjects (i.e. those subjects with some volitional
contraction in at least one lower limb muscle).
1156
B. Calancie et al.
will, or its discharge rate could be modulated if spontaneously
active) was also responsive to tibial nerve stimulation.
persons with motor-incomplete injury, contradicting one
conclusion from an earlier study (Calancie 1991); and (ii)
these re¯exes are not evident at the acute stage, but take many
months to emerge following spinal cord injury.
Discussion
In previous studies, we showed that interlimb re¯exes are
commonly seen in persons with chronic cervical spinal cord
injury (Calancie, 1991; Calancie et al., 1996). In the present
study, we now show that: (i) these re¯exes are also evident in
Table 2 Average tendon response amplitudes (mV) for the
two groups of SCI subjects at the time of initial testing and
at the ®nal test session prior to the emergence of ILR
activity
Achilles initial
Achilles ®nal
Patellar initial
Patellar ®nal
Motor-complete
Motor-incomplete
0.23
0.87
0.05
0.17
0.76
2.81
0.46
0.70
6
6
6
6
0.41
1.41
0.13
0.18
6
6
6
6
0.86
3.45
0.73
0.64
Origins of interlimb re¯exes
We suggest that ILR activity is mediated by sprouting from
ascending afferent ®bres onto motor neurones of a particular
upper limb muscle partially denervated following a more
rostrally-placed spinal lesion. Either new contacts between
afferents and these motor neurones are made, or existing
contacts are dramatically strengthened through the sprouting
process. While absolute proof of such sprouting is lacking, we
believe that the aggregate of evidence summarized below
points to this as the most probable explanation.
Distribution of interlimb re¯exes
The nature and extent of ILR activity was often restricted to a
few individual motor units recruited in one or several upper
limb muscles. These singular activations, with highly reproducible latency, commonly resulted in visible twitches in the
muscle belly or the joint being acted upon by the muscle. This
form of recruitment is very different from that typically seen
in response to synchronized peripheral (i.e. Ia) or central (i.e.
Between three and ®ve taps were applied at each of the Achilles
and patellar tendons. The single largest response was used for
subsequent calculations. Differences between initial and ®nal
values were statistically signi®cant for the motor-complete
Achilles, motor-complete patellar and motor-incomplete Achilles
comparisons.
Table 3 Distribution of ILR activity. Incidence of EMG response in given left-side (L) or
right-side (R) muscle to stimulation of the left-side or right-side tibial nerve across all
subjects
Left-side stimulation
Muscle with positive ILR response
R
Biceps brachii
Triceps brachii
Wrist extensors (ECR)
Wrist ¯exors (FCR)
Thenar group (APB)
Hypothenar group (ADM)
LLR
LLR
LLLRRRR
RRRRRR
Right-side stimulation
LLL
RLL
RRLLLLLL
RRRRLLL
LLLLLLLL
For example, the right thenar group (APB) showed two instances in which stimulation of a right-side (i.e.
ipsilateral) tibial nerve resulted in short-latency recruitment and four instances of recruitment to left-side
(i.e. contralateral) stimulation.
Table 4 Earliest (minimum latency) and most prolonged (maximum latency) onset of tibial
nerve stimulus-evoked EMG in different upper limb muscles in subjects demonstrating
interlimb re¯exes
Muscle
Minimum latency (ms)
Maximum latency (ms)
Average (ms)
SD (ms)
Biceps brachii
Triceps brachii
ECR
FCR
APB
ADM
50.0
±
32.0
32.0
43.9
46.2
50.0
±
64.0
56.5
104.0
110.0
50.0
±
46.0
43.7
58.4
63.4
±
±
9.8
11.0
15.3
17.2
The mean and standard deviation are given when more than one observation in a given muscle group was
made.
Interlimb re¯exes after spinal cord injury
corticospinal tract) excitatory inputs (Jones et al., 1996),
whereby excitation is likely distributed widely throughout the
motor neurone pool. Functional movements of the hand or
forearm were never observed.
As shown in Table 1, upper limb muscles which responded
to tibial nerve stimulation always derived their primary level
of innervation from neurologic levels caudal to the rostralmost level of fracture. For example, the single instance of ILR
activity observed in the biceps brachii muscle was from a
subject with a C3 fracture (Subject 4; this subject required
ventilator support for 3 weeks following his injury). Because
the biceps brachii receives its primary innervation from the
®fth and sixth neurological levels of the spinal cord (Kendall
and McCreary, 1983), motor neurones to this muscle would
be expected to be at least partially denervated by the more
cranially-placed spinal injury.
Based on our ®ndings, the ability to contract an upper limb
muscle through volitional effort does not preclude the
presence of an ILR response to lower limb inputs in that
same muscle. We saw a number of cases in individuals with
both complete and incomplete SCI where an EMG interference pattern with voluntary effort could be produced in the
same muscle from which ILR activity was seen. However,
muscles with positive ILR responses to lower limb innocuous
stimulation were invariably weaker during attempted voluntary contractions. This suggests that a greater percentage of
motor neurones to such muscles had lost innervation from
axons originating cranial to the lesion, rendering their motor
neurones more susceptible to regenerative sprouting from
lower limb ascending ®rst- or second-order afferent ®bres.
Indirectly supporting this suggestion, we did not see in this
studyÐnor in other subjects with chronic SCI studied at other
timesÐa single instance in which the same motor unit
recruited by voluntary contraction was also activated by tibial
nerve stimulation or other types of lower limb sensory inputs
(Calancie, 1991; Calancie et al., 1996).
These observations argue that those upper limb motor units
being recruited through tibial nerve stimulation lie towards
the caudal range of the motor neurone pool for that muscle.
There is abundant evidence (albeit from lower- or hind-limb
studies) supporting the concept that the motor neurones
innervating a given muscle are distributed in a rostral-caudal
orientation spanning three or more neurologic segments
(Henneman and Mendell, 1981; Ungar-Sargon and
Goldberger, 1987; Phillips and Park, 1991). In the human
upper limb, Kendall and McCreary (1983) have compiled
®ndings from six groups indicating that in man, there is also
widespread spinal segmental distribution to the six upper
limb muscles studied herein as follows: biceps: C5±C7;
triceps: C6±T1; ECR: C5±C8; FCR: C6±C8; APB: C6±T1;
ADM: C7±T1.
Given the requirement for a distribution of motor neurones
in a given muscle's pool to account for both ILR and
voluntary activity, this explanation also requires that the
longitudinal extent of spinal cord injury be relatively
circumscribed, preserving substantial grey matter caudal to
1157
the injury epicentre (at least for relatively mild injuries
leading to ASIA categories C or D). Limited evidence from
human histopathology studies indeed shows that traumatic
injuries are often restricted in a longitudinal (i.e. cranial±
caudal) extent to distances of no more than 3±4 cm within the
spinal cord (Bedbrook, 1963; Bunge et al., 1993, 1997).
For sprouting to occur onto or expand across motor
neurones of the cervical enlargement, somatic and/or
dendritic surface area must presumably be vacated due to
degeneration of synaptic contacts originating from nerve
®bres axotomized at a point more rostral in the neuraxis (i.e.
the site of spinal cord injury), and new synaptic contacts
formed (Raisman and Field, 1973). This partial denervation is
most likely to occur with motor neurones innervating muscles
of the forearm and hand, whose source of motor innervation is
highly dependent upon the corticospinal tract (Phillips and
Porter, 1977; Kuypers, 1981; Palmer and Ashby, 1992; Maier
et al., 1998) or cervical propriospinal neurones subserving
fractionated movements (Pierrot-Deseilligny, 1996).
Conversely, proximal upper limb muscles, such as biceps
and triceps, would be less likely to undergo denervation due
to the lesser dependence of their motor neurones upon
corticospinal in¯uence (either directly or via propriospinal
neurones) and greater input from other pathways (Turton and
Lemon, 1999), which appear to rely to a greater extent upon
segmental interneurones rather than direct supraspinal projections (Lawrence and Kuypers, 1968). The present ®ndings
agree with both predictions. That is, ILR activity is most
commonly seen in hand intrinsic muscles, is less common in
forearm muscles and is only rarely encountered in proximal
muscles of the upper limb, even in cases in which these
proximal muscles receive innervation from motor neurones
located within the caudal-most portions of the cervical
enlargement, such as the triceps brachii (in which we have
never observed ILR activity).
Time of interlimb re¯ex emergence post-injury
Does the time course of ILR emergence following spinal cord
injury support the argument that regenerative sprouting is the
most likely explanation for the ILR activity described herein?
We believe the answer is `yes', for the following reasons.
First, many months were needed before ILR activity was
seen, whereas the time at which `latent' synapses have been
demonstrated following nerve injury in various models
ranges from several minutes to several days (Nelson et al.,
1979; Wall, 1988; Goshgarian et al., 1989; Brasil-Neto et al.,
1992). Note also that 50% of the subjects had spinal re¯exes
that were brisk and well-de®ned immediately following
injury (i.e. those with motor-incomplete SCI), yet these
subjects had no evidence of ILR activity for many months
beyond this time. The majority of the remaining, motorcomplete subjects had recovered spinal re¯ex excitability
even before the experimental session in which ILR activity
was ®rst noted. Thus, one cannot simply invoke `spinal
1158
B. Calancie et al.
shock' as a basis for the absent interlimb re¯exes until six or
more months post-injury.
Secondly, the rate of cellular degeneration following
human SCI is considerably slower than that associated with
the rat based on histopathologic ®ndings (Becerra et al., 1995;
Bunge et al., 1997). This protracted period of cellular
breakdown would thereby delay the availability of new
synaptic sites on affected motor neurones.
Thirdly, nerve growth factor associated sprouting in human
axons has been reported following traumatic spinal cord
injury (Wang et al., 1996). In a different study, ascending
afferent ®bres within the dorsal columns were found to persist
just below the level of a cervical injury and did not retract
from the injury locus (Quencer and Bunge, 1996). Sprouting
from these ascending sensory ®bres may account not only for
the ILR activity reported herein, but also for the development
of autonomic dysre¯exia in this population over a period of
months following SCI (Mathias et al., 1979; Weaver et al.,
1997; Krenz et al., 1999; Krassioukov et al., 1999; Teasell
et al., 2000). We are currently pursuing this line of
investigation.
Finally, of the more than 70 subjects with chronic, severe
injury to the cervical spinal cord examined for ILR activity in
our laboratory, only four have not shown such re¯ex
behaviour when examined appropriately. We believe the
absence of ILR activity in these four subjects is due to a
widespread loss of spinal cord grey matter and resultant
denervation of hand and forearm muscles. We have con®rmed this in three of these subjects by demonstrating a
complete absence of intrinsic hand muscle contraction
following intense (>50 mA) electrical stimulation of median
and ulnar nerves at the wrists bilaterally (results not shown).
In our experience, such extensive grey matter loss after
cervical SCI is rare (see Peckham et al., 1976).
Latency of interlimb re¯exes
In records presented in this paper and elsewhere (Calancie,
1991; Calancie et al., 1996) the ILR onset latency to tibial
nerve stimulation (at the knee) can be as short as 50 ms or
less, when measured at the intrinsic muscles of the hand (note
that this latency is relative to the onset of the ®rst stimulus
pulse of the brief train used). The conduction path for this
activity in a subject whose height was ~6 feet (as was true for
Subject 4) would be approximately 2 metres in length. Of
course, there must be a minimum of two synaptic delays in
this path prior to the onset of EMG, giving rise to a calculated
conduction velocity of the fastest ®bre(s) mediating this
activity of no less than 42 m/s. We use the term `no less'
because a high frequency 3- or 4-pulse train is often necessary
to elicit interlimb re¯exes, suggesting that the evoked
response is elicited through the effects of temporal summation. Were latencies to be counted only from the onset of the
penultimate pulse, the calculated maximum axonal conduction velocity from the example above would be at least 48 m/s.
Based on the brief response latencies shown in this study, it
seems clear that the initial components of interlimb re¯exes
are mediated by spinal-only pathways, consistent with the
hypothesis that interlimb re¯exes re¯ect altered spinal
circuitry. Figure 4 shows that while ILR activity at 50±
60 ms is likely entirely of spinal origin, activity beyond a
latency of ~110 ms conceivably includes both spinal (relatively slowly-conducting) and supraspinal activity.
Many of the properties of interlimb re¯exes suggest that
they are mediated, at least in part, by second-order cells with
widespread convergence from primary afferents. We showed
in an earlier study that ILR response properties to repeated
stimulation can show `wind-up' (Mendell, 1966; Calancie
et al., 1996); that is, enhanced output (intensity of discharge)
for a given input. We interpret our ®ndings that modest
changes in the stimulus rate can lead to consistent reduction
in ILR response latency (as shown in Fig. 1) as being
consistent with recruitment of more rapidly conducting
second-order afferents re¯ecting this state of heightened
response probability underlying wind-up. Reliance upon
these ®bres and ascending routes also helps explain the
wide range of response latencies for different muscles as
reported in Table 4. While the fastest conducting afferent and
efferent limbs must be responsible for the shortest latency
responses seen, a combination of slowing conducting
afferents (primary and secondary) and relatively slowlyconducting motor axons could easily give rise to ILR
latencies of 100 ms and beyond.
We also show that the addition of extra pulses in a train can
actually reduce the response latency of ILR activity. We
typically used an electrical stimulus to the tibial nerve of
suf®cient intensity to cause a clear plantar-going twitch, such
that we were likely stimulating a large number of nerve ®bres
(both efferent and primary afferents). The 4-pulse train in the
example shown (Fig. 2) caused recruitment at ~56 ms. We
suggest that afferents of relatively rapid conduction velocity
mediated this response, and that the fourth pulse in the
stimulus train actually caused the second-order cell (or cells)
to reach threshold and discharge an action potential. By
eliminating one of these inputs (i.e. using a 3-pulse instead of
a 4-pulse train), activity in these more rapidly conducting
afferents could contribute to the second-order cell's depolarization, but additional effects of more slowly conducting
afferents were necessary to actually reach threshold, accounting for the additional latency (~4.5 ms in Fig. 2).
Other studies
There have been other studies describing the in¯uence on
upper limb musculature of lower limb sensory inputs (Gassel
and Ott, 1973; Kearney and Chan, 1979, 1981; Delwaide and
Crenna, 1984). Several authors have shown that stimuli
similar in principal to those used in the present study can
cause modulation of upper limb re¯ex excitability at latencies
that suggest spinal action in normal subjects (Kearney and
Chan, 1979; Delwaide and Crenna, 1984). However, in these
Interlimb re¯exes after spinal cord injury
studies, the magnitude of effect is sub-threshold for direct
observation, requiring a conditioning test paradigm to
resolve; this is in sharp contrast to the ease with which ILR
activity can be elicited and directly observed in persons with a
history of cervical SCI. Moreover, the latencies of effects
reported in these other studies are well beyond those seen in
the present study once the conduction delay associated with
the efferent limb of the motor circuit is taken into consideration. At the very least, the ILR activity reported in this
paper may re¯ect pre-existing segmental connections whose
ef®cacy is dramatically altered (i.e. enhanced) as a consequence of the more rostrally-placed spinal cord injury.
Signi®cance
We believe these ®ndings demonstrate segmental reorganization in the human spinal cord caudal to the level of injury, in a
manner not previously reported in this population (other than
in our earlier publications). The prolonged timeframe argues
that this form of plasticity may continue for many months, or
even years, after human injury. Whether or not the ILR
activity seen in these subjects remains constant (as measured
through response distribution, latency and post-stimulus time
histogram analysis) through subsequent follow-up examinations in these same subjects will form the basis of a
subsequent report.
These observations of presumed human spinal cord
plasticity bolster arguments for repeated episodes of rehabilitation in humans with sub-acute or chronic SCI (i.e. >1 year
duration) and further argue that animal models of SCI and
therapeutic intervention should include prolonged (i.e.
>1 year) survival durations post-injury to more accurately
model the human condition.
Acknowledgements
We wish to thank Natalia Alexeeva for her helpful comments
related to this manuscript. This work was supported in part by
the National Institutes of Health (HD31240; NS36542), and
by The Miami Project to Cure Paralysis.
References
Alexeeva N, Broton JG, Suys S, Calancie B. Central cord syndrome
of cervical spinal cord injury: widespread changes in muscle
recruitment studied by voluntary contractions and transcranial
magnetic stimulation. Exp Neurol 1997; 148: 399±406.
Bach Y, Rita P. Central nervous system lesions: sprouting and
unmasking in rehabilitation. Arch Phys Med Rehabil 1981; 62:
413±7.
1159
Post MJ. MR-pathologic comparisons of Wallerian degeneration in
spinal cord injury. AJNR Am J Neuroradiol 1995; 16: 125±33.
Bedbrook GM. Some pertinent observations on the pathology of
traumatic spinal paralysis. Paraplegia 1963; 1: 215±27.
Brasil-Neto JP, Cohen LG, Pascual-Leone A, Jabir FK, Wall RT,
Hallett M. Rapid reversible modulation of human motor outputs
after transient deafferentation of the forearm: a study with
transcranial magnetic stimulation. Neurology 1992; 42: 1302±6.
Brown AG, Fyffe RE, Noble R, Rowe MJ. Effects of hind limb
nerve section on lumbosacral dorsal horn neurones in the cat. J
Physiol (Lond) 1984; 354: 375±94.
Bunge RP, Puckett WR, Becerra JL, Marcillo A, Quencer RM.
Observations on the pathology of human spinal cord injury. A
review and classi®cation of 22 new cases with details from a case of
chronic cord compression with extensive focal demyelination. Adv
Neurol 1993; 59: 75±89.
Bunge RP, Puckett WR, Hiester ED. Observations on the pathology
of several types of human spinal cord injury, with emphasis on the
astrocyte response in penetrating injuries. Adv Neurol 1997; 72:
303-15.
Calancie B. Interlimb re¯exes following cervical spinal cord injury
in man. Exp Brain Res 1991; 85: 458±69.
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 1994; 117: 1143±59.
Calancie B, Lutton S, Broton JG. Central nervous system plasticity
after spinal cord injury in man: interlimb re¯exes and the in¯uence
of cutaneous stimulation. Electroencephalogr Clin Neurophysiol
1996; 101: 304±15.
Calancie B, Molano M, Broton JG. Neural plasticity as revealed by
the natural progression of movement expressionÐboth voluntary
and involuntaryÐin humans after spinal cord injury. [Review].
Prog Brain Res 2000; 128: 71±88.
Calancie B, Molano M, Broton JG, Bean JA. Relationship between
EMG and muscle force after spinal cord injury. J Spinal Cord Med
2001; 24: 19±25.
Christensen MD, Hulsebosch CE. Chronic central pain after spinal
cord injury. [Review]. J Neurotrauma 1997; 14: 517±37.
Delwaide PJ, Crenna P. Cutaneous nerve stimulation and
motorneuronal excitability. II: evidence for non-segmental
in¯uences. J Neurol Neurosurg Psychiatry 1984; 47: 190±6.
Ditunno JF, Young W, Donovan WH, Creasey G. The international
standards booklet for neurologic and functional classi®cation of
spinal cord injury. Paraplegia 1994; 32: 70±80.
Gassel MM, Ott KH. Patterns of re¯ex excitability change after
widespread cutaneous stimulation in man. J Neurol Neurosurg
Psychiatry 1973; 36: 282±7.
Barbeau H, McCrea DA, O`Donovan MJ, Rossignol S, Grill WM,
Lemay MA. Tapping into spinal circuits to restore motor function.
[Review]. Brain Res Brain Res Rev 1999; 30: 27±51.
Goldberger ME, Murray M, Tessler A. Sprouting and regeneration
in the spinal cord: their roles in recovery of function after spinal
injury. In: Gorio A, editor. Neuroregeneration. New York: Raven
Press; 1993. p. 241±64.
Becerra JL, Puckett WR, Hiester ED, Quencer RM, Marcillo AE,
Goshgarian HG, Guth L. Demonstration of functionally ineffective
1160
B. Calancie et al.
synapses in the guinea pig spinal cord. Exp Neurol 1977; 57: 613±
21.
Mendell LM. Modi®ability of spinal synapses. [Review]. Physiol
Rev 1984; 64: 260±324.
Goshgarian HG, Yu XJ, Rafols JA. Neuronal and glial changes in
the rat phrenic nucleus occurring within h after spinal cord injury.
J Comp Neurol 1989; 284: 519±33.
Mendell LM, Munson JB, Arvanian VL. Neurotrophins and
synaptic plasticity in the mammalian spinal cord. [Review].
J Physiol (Lond) 2001; 533: 91±7.
Guth L. Axonal regeneration and functional plasticity in the central
nervous system. Exp Neurol 1974; 45: 606±54.
Nelson SG, Mendell LM. Enhancement in Ia-motorneuron
synaptic transmission caudal to chronic spinal cord transection.
J Neurophysiol 1979; 42: 642±54.
Harkema SJ, Hurley SL, Patel UK, Requejo PS, Dobkin BH,
Edgerton VR. Human lumbosacral spinal cord interprets loading
during stepping. J Neurophysiol 1997; 77: 797±811.
Henneman E, Mendell LM. Functional organization of motorneuron
pool and its inputs. In: Mountcastle VB, Brookhart JM, Brooks VB,
editors. Handbook of physiology, Sect. 1, Vol. 2, Pt. 1. Bethesda
(MD): American Physiological Society; 1981. p. 423±507.
Hiersemenzel LP, Curt A, Dietz V. From spinal shock to spasticity.
Neuronal adaptations to a spinal cord injury. Neurology 2000; 54:
1574±82.
Jones KE, Calancie B, Hall A, Bawa P. Comparison of peripheral Ia
and corticomotorneuronal composite EPSPs in human motor
neurones. Electroencephalogr Clin Neurophysiol 1996; 101: 431±7.
Kearney RE, Chan CW. Re¯ex response of human arm muscles to
cutaneous stimulation of the foot. Brain Res 1979; 170: 214±7.
Kearney RE, Chan CW. Interlimb re¯exes evoked in human arm
muscles by ankle displacement. Electroencephalogr Clin
Neurophysiol 1981; 52: 65±71.
Kendall FP, McCreary EK. Muscles. Testing and function. 3rd ed.
Baltimore: Williams & Wilkins; 1983.
Krassioukov AV, Bunge RP, Puckett WR, Bygrave MA. The
changes in human spinal sympathetic preganglionic neurones after
spinal cord injury. Spinal Cord 1999; 37: 6±13.
Krenz NR, Meakin SO, Krassioukov AV, Weaver LC. Neutralizing
intraspinal nerve growth factor blocks autonomic dysre¯exia caused
by spinal cord injury. J Neurosci 1999; 19: 7405±14.
Kuypers HGJM. Anatomy of the descending pathways. In:
Mountcastle VB, Brookhart JM, Brooks VB, editors. Handbook
of physiology, Sect. 1, Vol. 2, Pt. 1. Bethesda (MD): American
Physiological Society; 1981. p. 597±665.
Lawrence DG, Kuypers HG. The functional organization of the
motor system in the monkey. II. The effects of lesions of the
descending brain-stem pathways. Brain 1968; 91: 15±36.
Maier MA, Illert M, Kirkwood PA, Nielsen J, Lemon RN. Does a
C3-C4 propriospinal system transmit corticospinal excitation in the
primate? An investigation in the macaque monkey. J Physiol (Lond)
1998; 511: 191±212.
Mathias CJ, Christensen NJ, Frankel HL, Spalding JM.
Cardiovascular control in recently injured tetraplegics in spinal
shock. Q J Med 1979; 48: 273±87.
McCouch GP, Austin GM, Liu CN, Liu CY. Sprouting as a cause of
spasticity. J Neurophysiol 1958; 21: 205±16.
Mendell LM. Physiological properties of unmyelinated ®ber
projection to the spinal cord. Exp Neurol 1966; 16: 316±32.
Nelson SG, Collatos TC, Niechaj A, Mendell LM. Immediate
increase in Ia-motorneuron synaptic transmission caudal to spinal
cord transection. J Neurophysiol 1979; 42: 656±63.
Palmer E, Ashby P. Corticospinal projections to upper limb motor
neurones in humans. J Physiol (Lond) 1992; 448: 397±412.
Peckham PH, Mortimer JT, Marsolais EB. Upper and lower motor
neuron lesions in the upper extremity muscles of tetraplegics.
Paraplegia 1976; 14: 115±21.
Phillips LH 2nd, Park TS. Electrophysiologic mapping of the
segmental anatomy of the muscles of the lower extremity. Muscle
Nerve 1991; 14: 1213±8.
Phillips CG, Porter R. Corticospinal neurones: their role in
movement. London: Academic Press; 1977.
Pierrot-Deseilligny E. Transmission of the cortical command for
human voluntary movement through cervical propriospinal
premotor neurones. [Review]. Prog Neurobiol 1996; 48: 489±517.
Quencer RM, Bunge RP. The injured spinal cord: imaging,
histopathologic clinical correlates, and basic science approaches
to enhancing neural function after spinal cord injury. Spine 1996;
21: 2064±6.
Raisman G, Field PM. A quantitative investigation of the
development
of
collateral
reinnervation
after
partial
deafferentation of the septal nuclei. Brain Res 1973; 50: 241±64.
Romero MI, Rangappa N, Li L, Lightfoot E, Garry MG, Smith GM.
Extensive sprouting of sensory afferents and hyperalgesia induced
by conditional expression of nerve growth factor in the adult spinal
cord. J Neurosci 2000; 20: 4435±45.
Schwab ME, Bartholdi D. Degeneration and regeneration of axons
in the lesioned spinal cord. [Review]. Physiol Rev 1996; 76:
319±70.
Siddall PJ, Loeser JD. Pain following spinal cord injury. [Review].
Spinal Cord 2001; 39: 63±73.
Steward O. Reorganization of neuronal connections following
CNS trauma: principles and experimental paradigms. [Review].
J Neurotrauma 1989; 6: 99±152.
Teasell RW, Arnold MO, Krassioukov A, Delaney GA.
Cardiovascular consequences of loss of supraspinal control of the
sympathetic nervous system after spinal cord injury. [Review]. Arch
Phys Med Rehabil 2000; 81: 506±16.
Turton A, Lemon RN. The contribution of fast corticospinal input
to the voluntary activation of proximal muscles in normal subjects
and in stroke patients. Exp Brain Res 1999; 129: 559±72.
Ungar-Sargon J, Goldberger ME. Maintenance of speci®city by
Interlimb re¯exes after spinal cord injury
1161
sprouting and regenerating peripheral nerves. II. Variability after
lesions. Brain Res 1987; 407: 124±36.
suggest reorganization of synapses on spinal sympathetic neurones
after cord transection. Neuroscience 1997; 81: 535±51.
Wall PD. Recruitment of ineffective synapses after injury. In:
Waxman SG, editor. Functional recovery in neurological disease.
Advances in neurology, Vol. 47. New York: Raven Press; 1988.
p. 387±400.
Wernig A, MuÈller S, Nanassy A, Cagol E. Laufband therapy based
on `rules of spinal locomotion' is effective in spinal cord injured
persons. Eur J Neurosci 1995; 7: 823±9.
Wang ZH, Walter GF, Gerhard L. The expression of nerve growth
factor receptor on Schwann cells and the effect of these cells on the
regeneration of axons in traumatically injured human spinal cord.
Acta Neuropathol (Berl) 1996; 91: 180±4.
Weaver LC, Cassam AK, Krassioukov AV, Llewellyn-Smith IJ.
Changes in immunoreactivity for growth associated protein-43
Wolpaw JR, Tennissen AM. Activity-dependent spinal cord
plasticity in health and disease. [Review]. Annu Rev Neurosci
2001; 24: 807±43.
Received September 13, 2000. Revised December 13, 2001.
Accepted January 9, 2002