Development 99, 221-230 (1987)
Printed in Great Britain © The Company of Biologists Limited 1987
221
Reformation of the pattern of neuromuscular connections in the
regenerated axolotl hindlimb
NIGEL STEPHENS and NIGEL HOLDER
Department of Anatomy & Human Biology, King's College London (KQC), Strand, London WC2R 2LS, UK
Summary
Retrograde neuronal tracing with horseradish peroxidase (HRP) was used to determine the position in the
spinal cord of motor neurone pools innervating
muscles in the regenerated axolotl hindlimb. This
method allows a detailed analysis of the accuracy of
reformation of neuromuscular connections. The results show that regenerated distal limb muscles are
reinnervated by motor neurones in the same region of
the cord as those that innervate normal control distal
limb muscles but that proximal muscles are innervated
by a mixture of motor neurones in a normal position
and motor neurones in a region of the spinal cord that
normally supplies innervation to distal limb muscles.
This difference between the reinnervation of proximal
and distal limb muscles suggests that axons destined
for proximal muscles may not enter distal limb territory during reinnervation of the regenerated limb.
Introduction
has also come from experiments in which limb nerves
have been cut and misrouted and reinnervation
patterns investigated by electrophysiology (Grimm,
1971; Cass, Sutton & Mark, 1973; Cass & Mark,
1975). More recently, using anatomical axon-tracing
methods it has been shown in this type of experiment
that nerve fibres grow back to their correct targets
from the point of deviation along the normal pathway
(Holder, Mills & Tonge, 1982; Holder, Tonge &
Jesani, 1984).
Evidence for nerve^target interactions has come
from electrophysiological analysis following either
denervation of a muscle and establishment of a
foreign innervation or muscle transplantation. In the
first instance whether the foreign innervation is derived from an implanted foreign nerve (Dennis &
Yip, 1978; Holder et al. 1982) or from nerve sprouting
from an adjacent normally innervated muscle (Bennett & Raftos, 1977; Genat & Mark, 1977; Wigston,
1980), the result is the same: matched axons establish
synapses of higher quantal content than foreign axons
and after a period of competition will exclude foreign
axons from the muscle. Selective reinnervation also
occurs when a dually innervated muscle is transplanted with its anterior-posterior axis reversed
(Wigston, 1986).
In urodele amphibians, amputation of a limb is
followed by a remarkable process of regeneration
which yields a structure morphologically identical to
that cut off. The morphogenetic mechanisms that
reproduce the normal pattern of tissues in the regenerate have been studied extensively (Tank & Holder,
1981; Wallace, 1981), but little is known about what
controls axonal regrowth or the accuracy of neuromuscular connections in the regenerate. Experimental investigation of nerve regeneration in urodeles has
shown that two (not necessarily exclusive) mechanisms exist: (1) interactions between the regrowing
nerve fibre and its immediate environment that define
the pattern of nerve growth within the limb; and
(2) interactions between axons and their target
muscles that favour formation of correct connections.
Evidence for these two distinct mechanisms has
come from several different types of experiment. The
importance of the environment through which axons
grow in determining nerve pattern is shown by
experiments in which the detailed anatomy of the
nerve pattern has been assessed after nerve regeneration or complete limb amputation (Weiss & Walker,
1934; Piatt, 1957). Evidence for local pathway cues
Key words: axolotl, motor neurone, neuronal specificity,
limb regeneration, pattern formation.
222
N. Stephens and N. Holder
Related experiments in teleost fish, which, like
urodeles, are known to restore coordinated movements following axon regeneration, have raised the
issue of whether specific neuromuscular regeneration
truly occurs. The original experiments in these
animals, which involved transection of nerves innervating extrinsic eye muscles and behavioural observations, suggested that specific nerve regeneration
did occur (Mark, 1974, 1980; Sperry & Arora, 1965).
However, in a recent study Scherer (1986) demonstrated that fish eye muscles are not selectively
reinnervated with respect to motor pool positions of
the oculomotor nerve, a finding that confounds the
original interpretations of Sperry & Arora (1965)
which were based on specific nerve-target matching.
This recent reassessment, combined with the fact that
all previous electrophysiological and anatomical assessments of specific reinnervation in urodele limbs
measured the 'correctness' of innervation using peripheral nerves (named limb nerves or spinal roots) as
reference, prompted us to examine the accuracy of
selective reinnervation using a more rigorous assay.
We have already studied the positions of motor
neurone pools in the normal axolotl hindlimb
(Stephens & Holder, 1985), and in the present study
this map is used to assess the degree of accuracy of
reinnervation of four muscles of the hindlimb following complete limb regeneration. We demonstrate that
spinal cord motor neurone pool positions of regenerated distal limb muscles are essentially normal but
that proximal muscles are innervated by motor
neurones in a normal position as well as motor
neurones in an inappropriate region of the spinal
cord.
Materials and methods
Animals and surgery
Sixteen white axolotls, all siblings from one spawning in the
colony at King's College, were used for this experiment.
Each had been reared separately from the time of limb
outgrowth to ensure that all limbs were not regenerates at
the time of operation.
Animals of length ll-5-16-0cm were anaesthetized with
MS222 and had their right hindlimb amputated through the
proximal thigh. After retraction of the soft tissues the
protruding segment of the femur was trimmedflushwith the
other stump tissues. Animals were allowed to survive
113-249 days before analysis and were fed maximally (two
or three times a week) with raw heart during this period. At
analysis, animals had grown to 15-0-19-7cm in length.
The survival time was judged to be sufficient both for the
process of pattern formation (morphogenesis in the forelimb blastema is complete within 30 days in similar-sized
animals reared at 21°C (Tank, Carlson & Connelly, 1976))
and for growth of the regenerate (although in fact the
regenerated hindlimb only achieved an average of 85-5 % ±
Table 1.
Muscle
PIT
Iliotibialis
Extensor
digitorum
Flexor
digitonim
Number
of cases
Survival
time
(days)
Mean length of
regenerate limb
w.r.t. control
(%±S.E.M.)
4
4
4
219
113
230
87-6 ±1-8
76-3 ±2-3
91-2 ±3-2
4
249
86-7 ±4-3
1-8% S.E.M. of the length of the control hindlimb during the
period of this experiment). Details of survival time and limb
growth for each muscle studied are given in Table 1.
Horseradish peroxidase (HRP) histochemistry
Operating and histochemical techniques were the same as
those used to construct the normal map of motor neurone
pools (Stephens & Holder, 1985). The same schedule was
used for all animals; HRP was placed in the puboischiotibialis (PIT), iliotibialis, extensor digitorum or flexor
digitorum muscle in the regenerated (right) limb and in the
same muscle in the unoperated (left) limb which was used
as a control. This comparison is fully justified because, as
we have already demonstrated (Stephens & Holder, 1985)
animals that have the same spinal root patterns have
equivalent motor pools in the same relative positions in the
rostrocaudal axis. It is always the case that the spinal root
patterns in the left and right sides of one animal are the
same. This procedure also allows a direct comparison
between the motor neurone pool of the normal muscle and
that of the regenerated muscle without reference to the
spinal nerve landmarks, and hence allowed inclusion of
cases with 'nonstandard' spinal nerve patterns. (These
animals, which were from a different spawning to those
used to construct the normal map of motor neurone pools
(Stephens & Holder, 1985), had an apparent caudal shift in
spinal nerve contribution to the crural plexus. Here, the
standard spinal nerve pattern included an additional contribution from nerve 18 in 8/16 cases, only 5/16 cases had the
'normal' 15-16-17 pattern described previously. In the other
3/16 cases there was no contribution from nerve 15.)
The position offilledmotor neurones in the rostrocaudal
axis was assessed by assigning them to bins 500 fun long.
The position of these cells relative to the spinal roots of the
crural plexus was also noted and the distribution recorded
as a histogram. The position of labelled motor neurones
was also assessed in the transverse plane with respect to
mediolateral and dorsoventral axes. For this analysis the
ventral horn was divided into a three-by-three grid which
was used as a reference with which to record the position of
each labelled cell (see Fig. 2). If a labelled cell was found in
the white matter it was recorded as being in the nearest
square. This analysis allowed an assessment of the position
of all labelled cells in the transverse plane, as well as
calculation of a mean coordinate position for each motor
neurone pool studied. This mean coordinate position is
plotted in Fig. 2.
Axolotl nerve regeneration 223
Table 2.
Mean number of labelled
cells ±S.E.M.
Regenerate
Control
PIT
Iliotibialis*
Extensor digitorum
Flexor digitorum
43-5
22-0
41-3
30-3
Ectopic cells labelled in
regenerate pool ± S.E.M.
±4-9
±5-0
±110
±0-9
43-5
38-0
32-3
33-8
+ 7-2
±3-9
±2-4
±2-7
Mean no.
Mean %
4-0 ±1-8
5-3 ±1-5
1-5 ±1-3
0
7-4 ±3-4
16-2 ±5-4
4-5 + 3-9
0
Mean distance
between medians of
control and
— regenerate pools
fim± S.E.M.
750 ±233
1583 ±155
312-5 ± 103
187-5 ±104
* One iliotibialis regenerate motor neurone pool contained only two cells. The data from this case are not included here.
Table 3. Mean coordinate positions of motor neurone pools in the transverse axes of the ventral horn
Control limb
Muscle
Medial-lateral
Regenerated limb
Dorsal-ventral
Medial-lateral
Dorsal-ventral
PIT
1-23
1-19
1-24
1-45
(0-08)
(0-06)
(0-07)
(0-08)
2-15
1-83
1-99
2-13
(0-1)
(0-08)
(0-08)
(0-08)
1-20
1-29
1-37
1-00
IT
1-31 ( 0 1 )
1-17(0-06)
1-12 (0-12)
1-35 (0-14)
2-43
2-10
2-25
2-47
(0-15)
(0-1)
(0-23)
(0-12)
1-22 (0-06)
1-36 (0-1)
1-45 (0-09)
—
2-25 (0-09)
2-47 (0-1)
2-51 (0-08)
—
EXT. DIG.
1-76(0-07)
1-64 (0-08)
1-77 (0-15)
1-74 (0-05)
2-49
2-26
2-28
2-44
(0-06)
(0-07)
(0-12)
(0-05)
1-83 (0-09)
1-66(0-08)
1-6 (0-07)
1-66 (0-06)
2-51
2-26
2-19
2-31
(0-08)
(0-09)
(0-06)
(0-06)
FLEX.DIG.
1-22
1-53
1-35
1-60
2-25
2-20
2-13
1-75
(0-08)
(0-1)
(0-08)
(0-1)
1-29
1-43
1-51
1-73
1-73
1-96
2-26
1-83
(0-12)
(0-1)
(0-09)
(0-12)
(0-07)
(0-07)
(0-09)
(0-08)
(±S.E.)
Results
Structure of the regenerated limbs
The anatomy of each muscle was carefully examined
by dissection at the time of HRP application and after
perfusion fixation for HRP histochemistry. Regenerate muscles were indistinguishable from controls in
14/16 cases. In one case a PIT muscle had an
abnormal tendinous band on its anterior border and,
in the other case, an iliotibialis muscle was reduced in
size, although this reduction was not quantified. This
muscle was subsequently shown to be innervated by
only two motor neurones.
Control motor neurone pools
The position of labelled motor neurones in the
transverse plane was assessed with reference to a
simple grid; the mean coordinate positions for the
proximal and distal muscles are shown separately in
Fig. 2 (the standard errors for each mean are given in
Table 3). The locations of extensor and flexor motor
pools were compared and it was found that the mean
(0-06)
(0-06)
(0-06)
(0)
(0-07)
(0-09)
(0-07)
(0-09)
1-68
1-86
1-97
2-00
(0-09)
(0-06)
(0-07)
(0)
locations of the extensor muscles (IT and extensor
digitorum) were clearly more medial than the location of the mean positions of motor neurones
innervating the flexor muscles (PIT and flexor digitorum). The four mean positions of the motor pools
of each muscle studied were clustered close together
with no overlap between pools. The degree of scatter
in the transverse plane of labelled cells for each pool
was similar from animal to animal and is shown for
individual representative cases in Fig. 3.
The relative rostrocaudal distribution of motor
neurone pools for the four control muscles studied
was similar to that described for the normal map of
hindlimb motor neurone pools (Stephens & Holder,
1985). However, there was a caudal shift with respect
to the spinal nerve roots of all pools in the 8/16
animals with a contribution from nerve 18 to the
crural plexus. In these cases the motor pools still
retained their relative positions in the lumbosacral
spinal cord: the proximal muscles, PIT and iliotibialis
were rostral, and the distal muscles extensor digitorum and flexor digitorum were caudal (Fig. 1). For
224
N. Stephens and N. Holder
PIT
IT
ED
16
FD
17
18
Fig. 1. The rostrocaudal distributions of motor pools for the four muscles in regenerates (solid bars) and controls
(dashed bars). The distributions are shown relative to spinal roots 16, 17 and 18 and extend from the most rostral to the
most caudal cell in any one pool. Also shown is the mean of each distribution which is indicated by an arrow. Note that
the arrows in the distal muscles, flexor digitorum (FD) and extensor digitorum (ED) for any pair of regenerate and
control distributions are clearly closer together than they are in the proximal iliotibialis (IT) and puboischiotibialis (PIT)
muscles. This disparity in relative mean positions is more clearly shown in Fig. 4.
Axolotl nerve regeneration 225
example, in one animal there were equal contributions from nerves 17 and 18 to the plexus and a
smaller contribution from 16. In this case, the control
PIT distribution extended from just above nerve 16 to
caudal nerve 17, in other words rostral with respect to
the whole limb-moving spinal cord region.
There was also an overall reduction in the mean
number of motor neurones labelled for three of the
four distributions studied (see Table 2) compared
c
c
Fig. 2. An outline drawing of the grey matter of the
lumbar spinal cord with a superimposed grid in the
dorsoventral and mediolateral axes showing the mean
coordinate positions of each motor pool for controls (*)
and regenerates (A). (A) Proximal muscles; (B) distal
muscles; M, medial; L, lateral; D, dorsal; V, ventral.
with our previous study. The exceptions to this were
the four control flexor digitorum motor pools, which
contained a mean of 30-3 cells, a similar value to the
distributions studied in constructing the normal map,
which had a mean of 32-5 cells. On the other hand,
the control PIT, iliotibialis and extensor digitorum
pools all contained less cells; means of 43-5, 220 and
41-3 respectively (in the normal map values of 62-2,
28-8 and 68-8 were obtained, Stephens & Holder,
1985). However, the relative values are maintained
here, with PIT and extensor digitorum containing
approximately twice as many cells as iliotibialis. This
reduction in the number of cells labelled is probably
due to a reduction in peroxidase activity in the
batches of HRP-WGA conjugate used for this experiment (Sigma Chemical Company data).
Regenerate motor neurone pools
The motor neurone pools of the control muscles and
the regenerated muscles were compared using four
criteria: the mean number of cells in the pool; the
mean number of ectopic cells (i.e. 'errors') in the
regenerate pool (ectopic cells were arbitrarily classified as such in the regenerate pools if they were
>500 pan rostral or caudal to the contralateral control
distribution); the mean distance between the medians
of the pools of the control and regenerate; and the
position of cells in the transverse plane. These data
are summarized in Table 2. The characteristics of
control and regenerate motor pools will be described
for each of the four muscles studied.
Puboischiotibialis
The mean number of cells in the control and regenerate pools for this muscle was similar. In two of the
four cases a significant number (18-0 and 8-8%) of
cells in the regenerate pool was in an ectopic (caudal)
position in the spinal cord (Fig. 1). In three of the
four cases studied the median values of the regenerate distribution were caudal to the controls
(Figs 1, 4), although only to an obvious degree
(1500/xm) in one case. In the transverse plane the
mean coordinate positions of the motor neurone
pools in each case were very close to those of the
control side (Fig. 2), these forming a group clearly
dorsal and lateral to those of the IT motor pools.
Iliotibialis
The reinnervation of this muscle was the least precise
of the four by all of the criteria used. In fact, one of
the four regenerate muscles was virtually uninnervated (only two motor neurones at a position 1000 /im
caudal to the control distribution were labelled); the
muscle in this case was also considerably reduced in
size. Whether this was a cause or effect of the
hypoinnervation or was unrelated to it is unclear.
226
N. Stephens and N. Holder
IT
PIT
16
17
18
18
5
5
16
16
r
17
i
17
i
i
i
j
Fig. 3. The motor neurone pools of representative cases of a regenerated (left) and control (right) pools for each of the
muscles studied. Each dot represents an HRP-filled cell in the spinal cord. The outlines represent the grey matter and
are drawn every 1500 fun and the number of cells every 500^m is shown on the histogram relative to the spinal roots.
The degree of scatter in the transverse plane can be assessed with respect to the mean values shown in Fig. 2.
Axolotl nerve regeneration 211
Data from this animal are excluded from the calculations. In the other three cases, more cells (38-0 ±
3-9 S.E.M. as compared to 22-0 ±5-0 S.E.M.) innervated the regenerate muscle, and although a majority
of these cells appeared to be in a correct rostrocaudal
and transverse position (Figs 2, 3), a clear minority of
16-2% was in an abnormal, caudal position. The
medians of the three distributions were caudal to the
controls (Figs 2, 4), a mean shift of 1583 /xm. Interestingly, however, virtually all the labelled cells, including the caudal ectopic ones, were in a correct,
ventromedial, position in the transverse plane
(Figs 2, 3). In the transverse plane the means of each
motor pool lay close to those of the control muscles,
the cluster forming a distinct group ventral to that of
PIT and more medial than the distal flexor digitorum
(Fig. 2).
Extensor digitorum
In general, the fidelity of reinnervation of this muscle
was better than for the two proximal limb muscles.
However, in this case, rather fewer motor neurones
innervated the regenerates than the control muscles
(32-2 ± 2-4 S.E.M. versus 41-3 ± 11-0). Ectopic cells in
the rostrocaudal axis were found in only one of the
four regenerate motor pools - constituting 18-2 % of
the total number of cells in this pool (Fig. 1). The
medians of the distributions were within 500 /im of
one another in all cases. In the transverse plane the
means of the distributions of each pool lay close to
IT
rostral
PIT
Jl
1
, 1
1
. 1
1
^«
1
1
f
J
[
caudal
30
20
10
10
No. of cells
ED
20
30
FD
Fig. 4. Diagrammatic representation of the pooled rostrocaudal distributions for each muscle with the median values
indicated by arrows. These were compiled by superimposing the median values of the control distributions for each case
studied for each muscle. Note that the median regenerated pool position for IT is clearly more caudal than the control;
IT and PIT show caudal tails on regenerate distributions and the means and patterns of distribution for control and
regenerate pools for the distal muscles are almost identical.
228
N. Stephens and N. Holder
those of the control muscles in a ventromedial position within the ventral horn (Fig. 2).
Flexor digitorum
The precision of reinnervation of this regenerate was
the best of the four muscles studied. Similar numbers
of cells were in the control and regenerate motor
pools (30-2 ±0-9 as compared to 33-8 ±2-7). Also,
no rostrocaudal errors were found in the projection to
the regenerate, the medians of these distributions
were only separated by an average of 187-5/an. The
means of labelled motor neurone positions in the
transverse plane formed a close cluster with those in
the control motor neurone pool (Fig. 2). This group
was located in a dorsolateral position with respect to
that of extensor digitorum.
Discussion
The results presented here clearly demonstrate that
the great majority of motor neurones that reinnervate
muscles in a regenerated limb he within the bounds
of the normal motor neurone pool for each muscle.
Given that the technique that we have used allows
discussion only of cell populations and not the behaviour of individual cells we may conclude that limb
reinnervation is not random but shows marked selectivity. The process is not absolute, however, because
some motor neurones from inappropriate motor pool
positions are maintained for the duration of the
experiment, which is over 7 months in the case of the
extensor digitorum. In this regard it is interesting to
note that these incorrect connections are apparently
stably maintained whereas in experimental situations
where such synapses have been induced surgically
they tend to be eliminated within this period (Mark,
1980). The reason for this difference is unclear.
In addition to the rostrocaudal distribution described previously (Stephens & Holder, 1985) and
reiterated here the present study includes a detailed
assessment of the position of labelled cells in the
transverse plane of the ventral hom. The mean
coordinate positions of control motor pools with
respect to mediolateral and dorsoventral axes form
tight clusters which are discrete and nonoverlapping.
The flexor muscles (flexor digitorum and PIT) have
motor neurones in similar positions in the transverse
plane (Fig. 2) but these pools show little overlap
rostrocaudally (Figs 1, 3 and see Stephens & Holder,
1985). In contrast, the motor pools of the extensor
muscles (IT and extensor digitorum) show considerable overlap rostrocaudally (Figs 1, 3 and see
Stephens & Holder, 1985) but the mean coordinate
positions in the transverse plane are discrete (Fig. 2),
with the extensor digitorum pool lying more medial
and dorsal to that of IT. Thus, if mean positions of
pools are considered, the motor pool of each muscle
in the present study occupies a clear-cut area of the
ventral horn. It is clear, however, that the overlap
that exists between the populations of cells in any two
motor pools restricts any conclusions as to whether a
single neurone has reinnervated its appropriate
muscle.
Despite the limitations of the assay, the results
indicate a reinnervation of regenerated muscles that
is dependent on motor neurone position. This is most
clearly seen in the transverse plane where the mean
coordinate positions of regenerate motor pools are
within the area of the control distributions in all 15
cases (Fig. 2). The distributions of motor pools in the
rostrocaudal axis are also remarkably similar for the
distal control and regenerate muscles. Thus, although
variation is seen between animals (Fig. 1) the distributions for particular pools are almost identical for
extensor and flexor digitorum. This can be seen most
clearly if the mean distributions for control motor
pools are superimposed (Fig. 4) so that the relative
positions of the regenerated pools can be compared
directly. For both of the distal muscles studied the
rostrocaudal extent and pattern of distribution are
essentially identical. This is not the case for the
proximal muscles, however, where caudal tails are
seen on both distributions (Figs 1, 4). In addition to
the tail of ectopic caudal cells the IT regenerate
motor pool shows a clear overall caudal shift (Fig. 4)
suggesting that it is now innervated by a significant
number of motor neurones that formally innervated
distal muscles. A caudal shift in the innervation of IT
has also been noted by Wigston (1986) who used
electrophysiological methods to assess selective reinnervation with respect to spinal roots 15-18. This
shift occurs when the muscle is removed and replaced
and when it is removed and transplanted with its
anterior—posterior axis reversed. In the present study
the caudal shift in the median value is also reflected in
the increase in numbers of cells in the regenerated IT
motor pools (Table 2). By contrast, a shift in the
median of the distribution is not seen in PIT (mistakes in the innervation of this muscle appear to be
restricted to the few cells comprising the caudal tail
(Fig. 4)), and in this muscle the numbers of cells in
regenerate and control pools are the same.
The pattern of selective reinnervation in the regenerated limb gives some clues as to the mechanism
underlying the process. The aberrant distributions in
motor pool positions of the proximal muscles (particularly IT) suggest that axons bound for distal
muscles growing within the blastema or young regenerate may synapse too early (i.e. with proximal
muscles) and that such synapses are apparently
stable. Compatible with this explanation is the absence of any rostrally positioned 'mistakes' in the
Axolotl nerve regeneration
motor pools for the two distal muscles. If these
connections had been formed it would imply that
axons destined for proximal muscles had entered
distal, foreign, tissue in the blastema. The fact that
they do not is consistent with the notion that axons
may only grow through specific regions of limb tissue,
that is, tissues that would guide growing axons to
their correct targets (Stephens, Holder & Maden,
1985; Holder etal. 1982, 1984).
One further point to consider is the influence of
survival time on the pattern of neuromuscular connections. That is, it may not be coincidence that the
muscle that showed the greatest degree of nonselective reinnervation (IT) was also that examined after
the shortest period of regeneration (Table 1). However, recent results from our laboratory show that the
number of aberrant connections is not greater at
earlier stages of limb regeneration (Wilson & Holder,
unpublished results). This result also demonstrates
that the initial pattern of neuromuscular connections
is not random, further indicating that growing axons
are guided to their target muscles by environmental
cues.
The results presented here provide an interesting
comparison with those of Scherer (1986) on the
regeneration of fish extrinsic eye muscle innervation,
a system previously thought to have many similarities
with the urodele limb (Mark, 1974). Scherer's work
has clearly demonstrated at the cellular level that
there is no selectivity in the regeneration of these
connections despite the fact that specificity had been
strongly suggested by behavioural studies. By
contrast, the work reported here (using similar
retrograde neuronal tracing methods) does show
selectivity in neuromuscular reconnections. The
significance of this difference, whether it is between
muscles or species, remains to be seen.
In conclusion, the results demonstrate that the
distal limb muscles, extensor and flexor digitorum,
are selectively reinnervated by their appropriate
motor neurone pools following complete limb regeneration. Proximal muscles are not so accurately
reinnervated, although the proportion of mistakes is
markedly higher in IT than PIT. In both proximal
muscles the incorrect connections appear to be made
by neurones that previously innervated distal limb
muscles.
It is a pleasure to thank Steve Wilson for detailed
discussions about the assessment of motor pool position
and he and Jonathan Clarke for comments on the manuscript. The work was supported financially by the MRC.
229
References
M. & RAFTOS, J. (1977). The formation and
regression of synapses during reinnervation of axolotl
striated muscles. J. Physiol. 265, 261-275.
BENNETT,
CASS, D. T. & MARK, R. F. (1975). Reinnervation of
axolotl limbs 1. Motor nerves. Proc. R. Soc. Lond. B
190, 45-58.
CASS, D. T., SUTTON, T. & MARK, R. F. (1973).
Competition between nerves for functional connections
with axolotl muscles. Nature, Lond. 2A5, 201-203.
DENNIS, M. J. & YIP, J. W. (1978). Formation and
elimination of foreign synapses on adult salamander
muscle. J. Physiol. 274, 299-310.
GENAT, B. & MARK, R. F. (1977). Electrophysiological
experiments on the mechanism and accuracy of
neuromuscular specificity in the axolotl. Phil. Trans. R.
Soc. Lond. B 278, 335-347.
GRIMM, L. M. (1971). An evaluation of myotypic
respecification in axolotls. /. exp. Zool. 178, 479-496.
HOLDER, N., MILLS, J. & TONGE, D. A. (1982). Selective
reinnervation of skeletal muscle in the newt Triturus
cristatus. J. Physiol. 326, 371-384.
HOLDER, N., TONGE, D. A. & JESANI, P. (1984). Directed
regrowth of axons from a misrouted nerve to their
correct muscles in the limb of the adult newt. Proc. R.
Soc. Lond. B 22, 477-489.
MARK, R. F. (1974). Selective innervation of muscle. Br.
Med. Bull. 30, 122-126.
MARK, R. F. (1980). Synaptic repression at
neuromuscular junctions. Physiol. Rev. 60, 355-395.
PIATT, J. (1957). Studies on the problem of nerve pattern.
m . Innervation of the regenerated forelimb in
Amblystoma. J. exp. Zool. 136, 229-247.
SCHERER, S. (1986). Reinnervation of the extraocular
muscles in goldfish is nonselective. /. Neurosci. 6,
764-773.
SPERRY, R. W. & ARORA, L. (1965). Selectivity in the
regeneration of the oculomotor nerve in the cichlid
fish, Astronotus ocellatus. J. Embryol. exp. Morph. 14,
307-317.
STEPHENS, N. & HOLDER, N. (1985). A horseradish
peroxidase study of motorneuron pools of the forelimb
and hindlimb musculature of the axolotl. Proc. R. Soc.
Lond. B 224, 325-339.
STEPHENS, N., HOLDER, N. & MADEN, M. (1985).
Motorneuron pools innervating muscles in vitamin Atreated proximal-distal duplicate limbs in the axolotl.
Proc. R. Soc. Lond. B 224, 341-354.
TANK, P. W., CARLSON, B. M. & CONNELY, T. (1976). A
staging system for forelimb regeneration in the axolotl,
Ambystoma mexicanum. J. Morph. 150, 117-128.
TANK, P. W. & HOLDER, N. (1981). Pattern regulation in
the regenerating limbs of urodele amphibians. Q. Rev.
Biol. 56, 113-141.
WALLACE, H. (1981). Limb Regeneration. New York: J.
Wiley.
230
N. Stephens and N. Holder
P. & WALKER, R. (1934). Nerve pattern in
regenerated urodele limbs. Proc. Soc. exp. Biol. Med.
31, 810-812.
WIGSTON, D. (1980). Suppression of sprouted synapses in
axolotl muscles by transplanted nerves. /. Physiol. 307,
355-366.
WEISS,
D. (1986). Selective innervation of transplanted
limb muscles by regenerating motor axons in the
axolotl. J. Newosci. 6, 2757-2763.
WIGSTON,
{Accepted 3 October 1986)