/. Embryol. exp. Morph. Vol. 69, pp. 237-250,1982
237
Printed in Great Britain © Company of Biologists Limited 1982
A scanning electron microscope
study of the development of a peripheral sensory
neurite network
By ALAN ROBERTS AND J. S. H. TAYLOR
From the Department of Zoology, University of Bristol
SUMMARY
The formation of the sensory neurite plexus on the basal lamina of trunk skin in Xenopus
embryos has been examined using the scanning electron microscope. It is formed by RohorjBeard and extramedullary neurons which provide the first sensory innervation of the skifi.
By observing the distribution of growth cones on the inside surface of the skin of embryos lat
different ages, the development of the plexus has been followed and related to the development of sensitivity to sensory stimulation. The general features of the plexus are illustrated
using a photomontage taken at x 1100. Measurements on neurites from this, and of growth
cone orientations demonstrate a general ventral growth pattern with some small regional
variations. Interactions of neurites within the plexus are examined. Neurites meeting at
shallow angles tend to fasciculate, while those meeting at close to 90° tend to cross each oth$r.
Angles of incidence and separation of neurites show few angles less than 30°, which sugge$ts
that active adjustments occur after a growth cone meets or leaves another neurite. The
observations allow comparison of behaviour of growing neurites in vivo and in vitro. Our
evidence suggests that adhesion between growth cones and neurites is stronger than that
between growth cones and the basal lamina of the skin.
INTRODUCTION
Factors affecting the growth of peripheral neurites in vivo have been studied
by Harrison (1910) and Spiedel (1932, 1935). Constant reference to these two
classical papers on amphibians confirms their stature and the general lack of
information on the initial stages of growth of nerve pathways in whole embryos.
This contrasts with a wealth of information on cultured neurons (reviewed
recently by Johnston & Wessells, (1980)). We have used the scanning electron
microscope to study the first establishment of innervation fields in Xenopus
laevis embryos (Davies, Kitson & Roberts, in preparation; Taylor & Roberts,
in preparation). In this paper we examine the growth of the network, or plexus
of neurites formed on the inside surface of trunk skin by Rohon-Beard and
extramedullary cells (Harrison, 1910; Hughes, 1957; Roberts & Hayes, 1977).
While most of Spiedel's observations were on regenerating neurites in the tails
of older tadpoles, our results concern the first establishment of sensory innefvation of the trunk skin by unmyelinated neurites. We have been concerned
238
A. ROBERTS AND J. S. H. TAYLOR
mw^^m^^^^
MJ':
Figs. 1 and 2, general view of trunk and peeled skin. Fig. 1. Lateral view of the
rostral trunk of a stage-27 embryo after removal of the skin. Dorsal up, rostral to the
right, x 100.
Fig. 2. Inner surface of a skin piece removed from the rostral trunk of a stage-27
embryo. Dorsal up, rostral to the left, x 100.
throughout to evaluate factors which might influence the direction of growth
of these neurites and to obtain observations which would allow comparison of
our in vivo preparation with in vitro results. A preliminary note has been
published in which the form of Xenopus sensory neurite growth cones is described (Roberts, 1976).
METHODS
Embryos from stages 22 to 32 (Nieuwkoop & Faber, 1956) were removed
from their egg membranes and fixed in 4 % glutaraldehyde in 0-5 M cacodylate
buffer at pH 7-3 for 1-2 h. After fixation they were dissected in buffer using fine
tungsten needles. The skin was removed in a single sheet from each side of the
body. After dehydration in ethanol, specimens in dry acetone were critical-point
dried using CO2, mounted on stubs and sputter coated with gold. The skinned
trunk and inside surface of the skin were then examined in a Cambridge S4
stereoscan.
Figs. 1 and 2 show examples of the trunk and skin pieces at low magnification.
Both surfaces were examined at x 1100 or more to look for neurites and growth
cones. The preservation was generally good for the neuronal tissue. Some
tearing of neurites was noticed and was clear, since neurites were seen detached
from their substrate. Similar detachment of growth cone processes was also seen
Sensory neurite network
23
239
r
24
T i i
23 25
i
i i
27
i
i
29
i
32
i
33 Stage
25
1 mm
26
27
28
29
Growth cones
Sensitivity
Fig. 3. Distribution of growth cones at different stages (left-hand diagrams) in
relation to sensitivity to stimulation (right-hand diagrams). At each stage large dots
indicate the distribution of growth cones. Diagrams of stages 24, 26 and 28 include
data from more than one skin specimen. The graph plots the number of growth
cones on skin specimens at different stages. Stippled areas in right-hand diagrams
show areas of skin where gentle mechanical stimulation evokes movements (based
on Fig. 5 in Roberts & Smyth (1974)). Note that data were not available for
the skin of the tail region.
occasionally (e.g. Fig. 7). Pitting of the basal lamina of the skin is probably a
preservation artifact.
Measurements were made from photographic prints taken with the long axis
of the embryo horizontal on the microscope screen. Angles were measured to
±5° with a protractor. For one animal a photomontage of 173 pictures at
x 1100 was made of the trunk skin (Fig. 6). The results are based on examination
of over 25 embryos.
240
A. ROBERTS AND J. S. H. TAYLOR
Figs. 4 and 5. The plexus on the basal lamina. Fig. 4. Examples of fasciculation and
crossing of neurites at stage 29/30. There is no branching. A growth cone is arrowed,
x 2500.
Fig. 5. Branching (arrowhead), fasciculation and crossing of neurites at stage 28.
x2300.
Fig. 6. Neurite plexus on the basal lamina of trunk skin of a stage 26-27 embryo.
The location of the map of the plexus is shown on the diagram of the embryo. The
map is a reduced tracing from a photomontage. The dorsal and ventral edges of the
myotomes (dotted) were clear from indentations in the skin. The most rostral
myotome is the second post-otic. Further details in the text. Scale bar 300 /im.
Sensory neurite network
241
RESULTS
General pattern of development
Examination of the inside surface of the trunk skin showed that neurites and
growth cones first appear over the myotome region at stage 24 or 25. These
neurites belong to Rohon-Beard or extramedullary cells and arise from the
dorsal spinal cord (Roberts & Clarke, 1982; Taylor & Roberts, in preparation).
By mapping the distribution of the growth cones of these neurites at sequential
embryonic stages, the development of the neurite network under the skin was
followed (Fig. 3). After their first appearance over rostral myotomes, growth
cones were found more caudally, and ventrally as development proceeded.
They had reached the tail and ventral belly by stage 29 to 30. Outgrowth of
neurites to the skin is not synchronous at any one location. As the pioneers grow
away from the spinal cord later growth cones appear dorsally over the myor
tomes. The number of growth cones is increased by these additions to reach a
maximum at about stage 26 (inset in Fig. 3). By stage 29 there are no growth
cones over the myotomes and the total number has dropped. Outgrowth of
neurites from the cord must stop just before this stage.
Between stages 24 and 28 the ventral border of the growth cones in the mid
trunk moves about 500 /*m. These stages are 6 h apart, indicating a growth rate
of about 80 /*m per h at 22-24 °C. A comparison of the distribution of growth
cones and the areas of the body which when stimulated mechanically will initiate
movements (Fig. 3) shows that there is a marked delay of 3-5 h between the
arrival of growth cones under the skin and the establishment of sensory
function.
The details of the neurite network formed on the basal lamina of the skin can
be seen at magnifications from 1000 to 2500 (Figs 4 and 5). As Harrison (1910)
noted, neurites branch, cross each other, fasciculate and separate again to form
a dense plexus. A general view of one side of one embryo at stage 26/27 was
obtained by making a photomontage of pictures at x 1100 and tracing the
neurites revealed (Fig. 6). Since the neurites arise dorsally in the spinal cord and
most of the skin to be innervated lies ventral to the cord, it is clear that growth
must be primarily ventral in orientation. However, while showing such a
general trend the map (Fig. 6) also shows the complexity of the plexus. Neurites
lie in many orientations and there may be regional differences. Measures of
orientation will be considered in the next section.
On the basal lamina the main structural feature that neurites encounter is
other neurites. Figure 6 shows that features such as the indentations formed in
the skin by the myotomes do not noticeably affect the course of the neurites.
However, changes in the direction of growth results after fasciculation, separation of fasciculated neurites or branching (Figs. 4, 5). Factors which influence
branching remain unclear, but fasciculation appears to occur when neurites
approach at shallow angles. When they approach at right angles they usually
242
A. ROBERTS AND J. S. H. TAYLOR
8
V
rr. A
10
Sensory neurite network
243
Dorsal
Rostral
270'
90°
Ventral
Fig. 11. Orientation of growth cones at stages 24-33. Polar plots showing numbers
of growth cones in each 10° segment. The trunk skin was divided into three equal
longitudinal regions. Each of these was subdivided into dorsal (fin and dorsal
myotome), mid (ventral myotome and dorsal belly), and ventral (belly) skin giving
nine regions. The shortest segment length is one growth cone. 0° dorsal; 90° caudal;
180° ventral; 270° rostral.
Figs. 7-10. Growth cones interacting with neurites on the basal lamina. Fig. 7.
Example of a growth cone which meets a neurite at nearly 90° and crosses, rather
than fasciculating. Note mutual attachment as growth cone crosses neurite at a
varicosity, and some damaged micropodia on growth cone (arrowhead), x 2220.
Fig. 8. Example showing crossing and fasciculation. Two neurites (top middle)
fasciculate and then separate. A neurite from top left shows a simple growth cone
with one micropodium extending along the contacted neurite. This would probably
lead to fasciculation. The right-hand growth cone meets at 90° and crosses, x 2220.
Fig. 9. Two growth cones. The upper one meets at a shallow angle and will probably
fasciculate. Note attachment of neurites to substrate. * indicates artifact, x 2220.
Fig. 10. Growth cone fasciculating with a fine neurite (arrowhead) which it appears
to pull towards itself, x 2220.
244
A. ROBERTS AND J. S. H. TAYLOR
2-1
Rostral
-«—• I -'O—
r>Tl-
•8-
Caudal
OJ
Ventral
oo
0J
Longitud.
3020Density
10-
'8°
02
4
6
7
9
11 13 15 18 (S)
Postotic myotome
Fig. 12. Measures of orientation and density of neurites as a function of longitudinal
position. Samples were taken at mid-myotome level (closed circles) and dorsal belly
level (open circles). Caudal to the 19th post-otic myotome segmentation (S) is in
progress. Further details in the text. Stages 26-27.
cross each other (Wessells et al. 1980). That these patterns seen in established
parts of the plexus are primarily a result of the behaviour of growth cones rather
than later adjustment can be seen by examining the growing ends of neurites.
When growth cones approach a neurite at nearly 90° they usually cross (Figs. 7,
8) but if they make contact at shallower angles (Figs. 8, 9 and 10) fasciculation
results. Many observations of this type suggest that the layout of the plexus is
established by the way it grows. Later adjustment of the positions of neurites
is unlikely to be extensive since they appear to be anchored to the basal lamina
by many small attachments (see also Roberts, 1976). Measurements on the
occurrence of fasciculation are reported below.
Orientation of growth
In analysing orientation, measures were first made on growth cones. Since
the plexus map (Fig. 6) suggested that there could be regional variation in
orientation, growth cone angles were related to position on the trunk (Fig. 11).
These plots show an overall ventral orientation of the growing tips of the
neurites (306 angles, mean 186°). The angles are significantly non-uniform in
distribution (probability of uniformity less than 0-1 % by Rayleigh's test). The
Sensory neurite network
245
20 -i
Incidence
10"
10°
30
50°
70
90
Fig. 13. Distribution of angles of incidence and separation of neurites. For incidence
N = 68 and angles over 90° were excluded because of possible ambiguity about the
direction of growth. For separation N = 138, S.D. = 21-6, Mean 68°.
V1 test (Durand & Greenwood, 1958) also gives a better than 0-1 % probability
in testing for closeness to an a priori angle of 180°. All but one of the dorsaljy
oriented growth cones in the dorsal regions were in embryos older than stage 26.
These growth cones also have a characteristic morphology (Roberts & Taylor,
in preparation). Over the ventral myotomes and belly skin growth is broadly
ventral. No rostral or caudal tendencies are clear.
Orientation of neurites in the plexus was examined for the neurite map of
Fig. 6. A regular 10 cm grid was drawn on the original map at x 1100 magnification. An 8 cm diameter sampling circle was centred in turn on each junction of
the grid. The number of neurites crossing 1 cm spaced parallel lines in the circle
was counted: with the lines longitudinal (XA); turned 45° clockwise (XB), and
turned 45° anticlockwise (Xc) (see Fig. 6). The following ratios were then
calculated:
7 to indicate neurite density
> 1 if neurites are oriented rostrally
< 1 if neurites are oriented caudally
XA/(X
> 1 if neurites are oriented ventrally
< 1 if neurites are oriented longitudinally
It was assumed, on the basis of the growth cone orientation measures, that
growth in the midmyotome and dorsal belly regions was in a broadly ventral
direction. These ratios are related to longitudinal position in Fig. 12. The
density of neurites falls caudally, making the orientation measure insignificant
behind the 15th myotome. Rostral to the 6th myotome there is a significant
246
A. ROBERTS AND J. S. H. TAYLOR
lOOn
Angle
Fig. 14. Relationship between the angle of incidence of two neurites and whether
the neurites fasciculate or cross. The numbers of measurements at each angle are
indicated. The line through the points is based on a weighted linear regression
analysis (slope0-9128 ±0182, intercept 0-6116± 10-4, r = 77-5 %,F (1,66) = 99.12,
N = 68) and the dashed lines indicate the 95 % confidence limits.
rostral orientation of neurites (Binomial test 99 % confidence). Caudal to the
15th myotome neurites may have a longitudinal tendency, but rostral to this
the ventral orientation is clear (Biomial test 99 % confidence).
Neurite interactions
The neurite map (Fig. 6) was used to evaluate the behaviour of neurites when
they meet and separate. Random behaviour would suggest weak interactions,
but our qualitative observations led us to expect mutual attraction between
neurites and growth cones. The direction of growth must be known to allow
distinction of incidence from separation, so we have only used the area of skin
ventral to the dorsal edge of the myotomes where the neurites arise (see Fig. 6
and Taylor & Roberts, in preparation). Even in this area measurements were
excluded if there was any uncertainty about direction of growth. Angles of
neurite incidence and separation were measured (Fig. 13) and after incidence,
the number of neurites crossing or fasciculating was recorded (Fig. 14).
If there was no active interaction between neurites one might expect angles of
incidence to be randomly distributed, and Figure 13 shows that from 30° to 90°
this seems a reasonable expectation (x2 test). However, there is a significant lack
of shallow angles of incidence, which is particularly surprising as we have shown
that the neurites show a general ventral orientation. This might have resulted
in more, rather than less shallow angles of incidence. Active interaction of
neurites is therefore implicated. After meeting, the data of Fig. 14 shows that
Sensory neurite network
247
shallow angles of incidence lead to fasciculation while steeper angles result in
neurites crossing. The final measurements were on angles of separation of
neurites. The majority of cases were defasciculations but branching was npt
excluded, since in both cases neurites have to separate. The mean angle of separation was 68° and the angles were normally distributed with few angles less than
30° or greater than 110° (Fig. 13). Again, the absence of shallow angles suggests
active interaction between neurites.
DISCUSSION
The present SEM observations on fixed preparations of normal embryos
show neurite structure and behaviour very similar to that seen in earlier studies
on cultured frog neurites (Harrison, 1910) and on regenerating unmyelinated
neurites in older tree frog larvae (Spiedel, 1933). Later studies on a wide range
of cultured neurons show strikingly similar behaviour in the growing neurites.
These similarities can give us confidence in all these methods of observation.
The rate of neurite growth (80 /tm h" 1 at 22-24 °C) in the Xenopus embryos is
faster than in the other amphibians (up to 56 /*m h" 1 (Harrison, 1910); 30-40 /im
h" 1 (Spiedel, 1932)). This could result partly from expansion of skin behind
growth cones. The delay between arrival of growth cones under the skin and the
establishment of sensory function (3-5 h) seems long. At present we have no
satisfactory method to follow the growth between the skin cells. This problem
also makes factors influencing branching hard to assess since many branches go
through the basal lamina into the skin and would be hard to detect with our
SEM technique. It is in fact difficult to distinguish between branching and the
separation of two intimately joined neurites (Fig. 5).
Orientation
It seems likely that the early overall ventral orientation of growth results
from the way the neurites emerge from the spinal cord and then grow over the
myotomes whose curvature directs the neurites ventrally (Taylor & Roberts, in
preparation). The later phase of dorsal growth (Fig. 11 and Roberts & Taylor
in preparation) could depend on the changed morphology of the cord, myotomes
and skin by stage 27, but more observations are required to resolve this. Small
regional differences in orientation could result from towing of neurites by
relative movements of the skin and myotomes. The neurites on the basal lamina
of the skin have a general ventral orientation but this is far from strict, suggesting fairly weak determining factors rather than a single, powerful orienting
force. It seems plausible at present to suggest that the innervation pattern
(Roberts & Hayes, 1977) is determined by gross morphological features coupled
with the fact that each sensory Rohon-Beard cell has some limit to the area of
skin that it can grow to innervate.
248
A. ROBERTS AND J. S. H. TAYLOR
Interaction of neurites
The factors which determine whether neurites cross or fasciculate when they
meet has been a major concern in this study. It has allowed close comparisons
to be made between in vivo and in vitro observations. Before attempting to interpret our observations we need to consider the relevant properties of the neurites
and growth cones (reviewed by Johnston & Wessels, 1980).
(1) Growth cones and neurites in some way adhere to the substrate and in
the case of neurites on the basal lamina this may be by small branchlets (Figs. 710; Roberts, 1976).
(2) Behind growth cones the neurite can remain active, still capable of extending filopodia for some distance (Figs. 7-10; Roberts, 1976; Roberts &
Taylor, in preparation).
(3) Growth cones grow over each other and other neurites to which they
appear to adhere (Harrison, 1910; Spiedel, 1933;Nakai, 1960; Nakajima, 1965;
Roberts, 1976; Wessells et al. 1980, Figs. 7-10). In Xenopus we suggest that
other growth cones and neurites are more adhesive than the basal lamina.
(4) Growth cones, their processes and nearby neurites generate tension
between points which are anchored by adhesion (Bray, 1979; Nakai, 1960).
On the basis of this list of properties or assumptions, we can now interpret
our observations on neurite interactions.
Crossing occurs more as the angle of incidence approaches normality (Fig. 13)
At normal incidence the neurite provides a narrow area of greater adhesion
equally distributed on either side of the advancing growth cone. Tension will be
equally distributed and there is, therefore, no tendency for the growth cone to
change course and fasciculate. Fasciculation occurs with shallower angles of
incidence as the adhesive surface of the neurite becomes more unequal in its
effect on the growth cone. Filopodia will extend preferentially along the adhesive surface of the neurite (C. F. Letourneau, 1975). Fasciculation can occur
either on initial contact of the growth cone and neurite, or the growth cone
can cross and then be pulled back to fasciculate by filopodia contacting the neurite
after crossing. Two processes could contribute to the absence of shallow angles
of incidence of fasciculating neurites (Fig. 14). If the neurite is not firmly
anchored at the point where it is contacted, the growth cone could pull on it.
After contact the active region behind the growth cone could pull further,
zipping-up the two neurites into closer contact (Fig. 15). Such processes may
be occurring in Figs. 9 and 10 and have been illustrated by Nakai (1960) and
Spiedel (1933) who also shows that crossings of neurites can be transformed
into short fasciculations by the second of these two processes.
Separation of neurites implies that the adhesion offered by the basal lamina is
not much less than that of the neurite surface. Clearly the available area of basal
lamina is larger, so if a sufficient number of growth cone processes contact the
basal lamina, the growth cone can escape provided the angle of separation is
Sensory neurite network
249
Fig. 15. Interactions of neurites and growth cones, (a) Fasciculation: (1) contact,
(2) growing neurite pulls static neurite, (3) area of attachment is increased to change
apparent site of incidence and to widen the angle of incidence, (b) Separation of a
growth cone which then (3) pulls the neurite to increase the apparent angle of
separation. Points of attachment are shown as little branchlets.
sufficient. As in the case of incidence, angles of separation can be increased by
pulling (Fig. 15 b) and zipping-up of neurites.
It seems that differential adhesion could, therefore, account for most of Our
observations. Rutishauser, Gall & Edelman (1978) have shown in vitro that
chick ganglion cells fasciculate less when cultures are on a more adhesive substrate. It will be of interest to see whether, in more natural, whole embryo situations, the degree of fasciculation is dependent on (a) the adhesion of the substrate,
(b) the shape of the substrate (a narrow tube would provide less chance Of a
growth cone escaping fasciculation), (c) numbers or density of neurites (more
neurites offer a larger adhesive surface). Perhaps these factors alone provide
sufficient variables to account for the variety of behaviour seen in growing
neurites, like those considered here which come from Rohon-Beard and extramedullary neurons. Centrally, these form a compact dorsal sensory tract
(Roberts & Clarke, 1982). On emerging from the cord their neurites form
bundles which when they reach the skin break up to form the plexus we have
described (Taylor & Roberts, in preparation). Such apparently complex behaviour could be based on a few limited rules. For example, Katz & Lasek (1979)
have described how retinal ganglion cell axons from eye primordia transplanted
to the tail of Xenopus embryos grow axons to ascend in the spinal sensory tracts
to the brain. This could be entirely dependent on the ganglion cell axons fasciculating with Rohon-Beard neurites which they then followed back to the spinal
cord and up to the brain.
We thank: L. Balch for making the photomontage of the plexus; B. Porter, S. Martin and
K.S.Williams for technical assistance; Drs J. Rayner and G. M. Jarman for statistical
advice; Professor P. Haggett and N. French for discussions on measurements; the MRC for
financial support.
REFERENCES
BRAY, D. (1979). Mechanical tension produced by nerve cells in tissue culture. /. Cell $ci. 37,
391-410.
DURAND, D. & GREENWOOD, J. A. (1958). Modifications of the Rayleigh test for uniformity
in analysis of 2 dimensional orientation data. /. Geol. 66, 229-238.
250
A. ROBERTS AND J. S. H. TAYLOR
R. G. (1910). The outgrowth of the nerve fibre as a mode of protoplasmic movement. /. exp. Zool. 9, 787-848.
HUGHES, A. F. (1957). The development of the primary sensory system in Xenopus laevis
(Daudin). /. Anat., Lond. 91, 323-338.
KATZ, M. J. & LASEK, R. J. (1979). Substrate pathways which guide growing axons in
Xenopus embryos. J. comp. Neurol. 183, 817-832.
LETOURNEAU, P. C. (1975). Cell-to-substratum adhesion and guidance of axonal elongation.
DevlBiol. 44, 92-101.
JOHNSTON, R. N. & WESSELLS, N. K. (1980). Regulation of the elongating nerve fibre. Curr.
Top. Dev. Biol. 16, 165-206.
NAKAI, J. (1960). Studies on the mechanism determining the course of nerve fibers in tissue
culture. II. The mechanism of fasciculation. Z. Zellforsch. mikrosk. Anat. 52, 427-449.
NIEUWKOOP, P. D. & FABER, J. (1956). Normal Tables o/Xenopus laevis (Daudin). Amsterdam:
North-Holland.
NAKAJIMA, S. (1965). Selectivity in fasciculation of nervefibersin vitro. J. comp. Neurol. 152,
193-204.
ROBERTS, A. (1976). Neuronal growth cones in an amphibian embryo. Brain Res. 118, 526530.
ROBERTS, A. & CLARKE, J. W. D. (1982). The neuronatomy of an amphibian embryo spinal
cord. Phil. Trans. R. Soc. Lond. B 296, 195-212.
ROBERTS, A. & HAYES, B. P. (1977). The anatomy and function of 'free' nerve endings in an
amphibian skin sensory system. Proc. R. Soc. Lond. B 196, 415-429.
ROBERTS, A. & SMYTH, D. (1974). The development of a dual touch sensory system in embryos
of the amphibian Xenopus laevis. J. comp. Physiol. 88, 31-42.
RUTISHAUSER, U., GALL, W. E. & EDELMAN, G. M. (1978). Adhesion among neural cells
of the chick embryo. IV. Role of the cell surface molecule CAM in the formation of neurite
bundles in cultures of spinal ganglia. /. Cell Biol. 79, 382-393.
SPIEDEL, C. C. (1932). Studies of living nerves. I. The movements of individual sheath cells
and nerve sprouts correlated with the process of myelin-sheath formation in amphibian
larvae. /. exp. Zool. 61, 279-331.
SPEIDEL, C. C. (1935). Studies of living nerves. II. Activities of ameboid growth cones, sheath
cells and myelin segments, as revealed by prolonged observation of individual nerve fibers
in frog tadpoles. Am. J. Anat. 52, 1-79.
HARRISON,
WESSELLS, N. K., LETOURNEAU, P. C , NUTTAL, R. D., LUDUENA-ANDERSON, M. & GADUSCHEK, J. (1980). Responses to cell contacts between growth cones, neurites and ganglionic
non-neuronal cells. /. Neurocytol. 9, 647-664.
(Received 12 November 1981, revised 12 January 1982)