/ . Embryol exp. Morph. Vol. 63, pp. 29-51, 1981
Printed in Great Britain © Company of Biologists Limited 1981
29
Significance of cell-to-cell contacts for the
directional movement of neural crest cells
within a hydrated collagen lattice
By EDWARD M. DAVIS AND J. P. TRINKAUS1
From the Department of Biology, Yale University, New Haven
SUMMARY
Neural tubes whose neural crest had just begun migration were isolated from stage-14
chick embryos, cleaned with 01 % trypsin, and cultured in transparent hydrated collagen
lattices (HCL) in an effort to stimulate in part the three-dimensional environment through
which neural crest cells migrate in situ, in the embryo. The concentration of collagen in the
lattices varied from 50 /tg/ml to 390 /*g/ml. The mode of movement and contact behaviour
of neural crest cells migrating from the neural tube under these conditions were recorded
directly with time-lapse cinemicrography. Both their shape and their rate of translocation
were dependent on the concentration of collagen in the HCL. In low concentrations (50 fig/
ml to 105 /ig/ml), neural crest cells have elongate spindle shapes and translocate at an average
rate of 1 /tm/min, whereas in high concentrations (190/tg/ml to 390/tg/ml), their shape is
rounded, and they translocate at an average rate of only 0-5 /tm/min. Neural crest cells
migrate from neural tubes in these preparations principally in loose clusters, with a few
single cells in the lead. The cells in these groups display leading-to-trailing edge adhesions
and form tongues or streams of cells directed away from the neural tube. The paths of
migration of both individual cells and groups of cells are aligned with the collagenfibrilsof
the HCL, which radiate from the neural tube. The classical visible characteristic of contact
inhibition of movement, change in direction of cell movement after contact with other! cells,
was not observed; neither the rate of translocation nor the time spent migrating away from
the tube is dependent on the number of contacts between cells. It is concluded that the
directional movement of neural crest cells in HCL cultures does not depend on contact
inhibition of movement.
INTRODUCTION
Neural crest cells migrate from the neural tube during embryogenesis in
a highly directional manner (Horstadius, 1950; Weston, 1970; Noden, 1978).
Evidence presented by Weston & Butler (1966), Le Douarin & Teillet (1974) and
Noden (1975) indicates that the environment through which these cells pass
plays a crucial role in this directionality. Lofberg (1976) and Ebendal (1977)
have suggested that within this environment it is the extracellular matrix that
influences directional movement by a kind of contact guidance (Weiss, 1961).
However, since alignment of the substratum in and of itself can only give such
moving cells orientation, not directionality (Trinkaus, 1976), other factors must
1
Author's address: Department of Biology, Yale University, New Haven, Connecticut
06520, U.S.A. Please address reprint requests to J.P.T.
2
E M Bi 6 3
30
E. M. DAVIS AND J. P. TRINKAUS
be involved. Tosney (1978) and Noden (1978) suggest that one such factor could
be contact inhibition of cell movement (Abercrombie & Heaysman, 1966). If
neural crest cells show contact inhibition, this, combined with an aligned substratum, could provide the directionality required, at least during the initial
phases of their migration, (see Trinkaus, 1980).
To determine if this is so, it is necessary to observe the movement of these
cells directly as they leave the neural crest and migrate into the extracellular
matrix of the embryo. Unfortunately, however, the favoured materials for
studying the migratory behaviour of neural crest cells, amphibian and chick
embryos, are so opaque that direct observation of cell movement in situ is
precluded. We therefore deemed it necessary to construct a three-dimensional
quasi in vivo environment in vitro that would at the same time encourage cell
movement and be of sufficient transparency to permit direct observation of it.
Elsdale & Bard (1972), have provided just such an environment in the hydrated
collagen lattice (HCL) and Bard & Hay (1975) have shown that the form and
movement of one kind of cell in an HCL in vitro - chick corneal mesenchyme is remarkably similar to the form and movement of these same cells in the
cornea in vivo. And Yang et al. (1979) have recently confirmed these favourable
features by showing that epithelial cells of a primary mammary tumor maintain
sustained growth and three-dimensional organization in such a gel. In consequence, we decided to study the directional movement and social interaction
of neural crest cells in such hydrated collagen lattices in an effort to observe
whether contact inhibition of movement is at work and, if so, is related to their
directional movement. Cell contact behaviour was recorded with time-lapse
cinemicrography and the films thus obtained provided the basis for our analyses.
MATERIALS AND METHODS
Preparation of collagen for hydrated collagen lattice
Rat tail collagen was isolated and the collagen concentration estimated, as
previously described (Davis, 1980). Hydrated collagen lattices, as described by
Elsdale & Bard (1972), containing 26/*g/ml to 394/Mg/ml of collagen, were
prepared by mixing equal volumes of collagen stock solution and twice concentrated growth medium, adjusting the pH to 7-4 with 1 N-NaOH, and diluting
with normal strength growth medium (Delbecco's Modified Eagle's Medium
with 10 % fetal calf serum, 100 U/ml Penicillin, 0-25 mcg/ml fungizone, and
100 mcg/ml streptomycin) to the desired concentration of collagen.
Primary cultures of neural crest cells
Neural tubes, 15 to 20 somites long, were isolated from the seventh to the
most posterior somite of stage-14 (Hamburger & Hamilton, 1951) White Leghorn chick embryos (Spafas, Norwich, Conn.) and stripped of clinging mesenchyme cells by digestion with 0-1 % trypsin (Trypsin 1:250, Gibco) (Davis,
Locomotion of neural crest cells
31
i
1980). At this stage, neural crest cells have begun their migration. These neural
tubes, with their neural crest, were then suspended in an HCL and incubated,
as previously described (Davis, 1980).
Scanning electron microscopic examination of trypsin-treated neural tubes
Freshly isolated, trypsin-treated neural tubes that had not been cultured in
HCL cultures were fixed overnight in 2 % glutaraldehyde in PBS, postfixed in
1 % osmium buffered with PBS, dehydrated through a series of ethanol dilutions,
and dried in a Sorvall critical-point drier, according to Anderson (1951). Tliey
were then coated with 20 nm of gold, using a rotary shadower (Ladd), and
examined with an ETEC Scanner operating at 10 kV.
Time-lapse cinemicrography
Time-lapse films of migrating neural crest cells were made as previously described (Davis, 1980), except that the interval between frames was 5 or 7-5
seconds. Filming was begun about 12 h after culturing.
Polarization microscopy
Polarization microscopy was used to examine the orientation of the collagen
fibrils of the HCL. A 35 mm camera was mounted to a Zeiss Universal microscope modified for polarization optics with a Polanochromatic 16x/0-035
objective and 5 x ocular. A neural tube cultured in an HCL was placed on
a graduated rotating stage and the analyzer was adjusted to 86-8°. Photographs
of the negative and positive retardation images were taken for every 30* of
stage rotation with Kodak SO 115 film and were developed in HC110 developer
(Dilution D for 8 min), fixed with Kodak rapid fixer, and processed normally.
Rate of translocation
Analysis of the rate of translocation that used longer intervals than frame*byframe analysis would facilitate the analysis of larger amounts of data. Increasing
the interval, however, might also obscure changes in minute-to-minute activity.
To test for this, the rates of translocation were determined both frame-by-frame
and at 5 min intervals, using photographic prints from the films (see Tabl$ 1).
Since the rates of translocation estimated with the two different techniques were
nearly the same, it appears that increasing the time intervals for analysis to 5 min
is approximately as accurate a method for determining rate of movement as is
frame-by-frame analysis. Accordingly, all subsequent determinations of rate of
cell movement (Tables 2, 3, 5, 7, 8, 9, 10) are based on observations made at
5 min intervals.
Evidence that the cells studied are derived from the neural crest
Since our interest was solely in the emigration of neural crest cells, we had to
be certain that the cultured neural tubes were free of contaminating sorfritic
2-fi
32
E. M. DAVIS AND J. P. TRINKAUS
Table 1. Shape and locomotory behaviour of neural crest cells in HCL
cultures containing different concentrations of collagen
Rate of translocation (/fm/min)
(^g/ml)
50
80
105
130
200
260
390
Length
Width
4-4±l-7(22)
3-9±l-6 (32)
5-2±l-8 (26)
4-0±l-5(26)
2-0+1-1 (24)
2-3±l-0(23)
1-7 ±0-4 (25)
A
Frame-by-frame
1-02 ±0-60 (10)
O-94±O-33(15)
MO±O-86 (8)
0-96±0-19(10)
0-50±012 (6)
0-38 + 019 (7)
0-42 + 0-19 (12)
5-minute interval
102 ±0-69 (9) (335)
1-08 + 0-71 (16) (575)
l-05±0-70(12)(520)
l-14±0-74(10)(357)
0-53 ±0-38 (8) (394)
0-52 + 0-31 (7) (211)
0-51+0-31 (12) (438)
The average length-to-width ratios of neural crest cells translocating within HCL cultures
containing different concentrations of collagen were determined by projecting the films onto
a screen. Measurements were taken every 5 min and at the end of each film the ratios were
averaged. Width was measured where the cell was apparently thickest. The numbers in
parentheses equal the number of cells measured during the filming period of approximately
360 min.
Rates of translocation were determined by projecting films onto a screen, tracing the paths
of migration of the centre (midpoint between the leading and trailing edge) of translocating
cells frame by frame and dividing the distance travelled by the time in minutes. The numbers
in parentheses equal the number of cells observed. The time of observation varied from 200
to 360 min.
Rates of translocation were also determined by making prints from time-lapse films at
5 min intervals. The progress of the cells was measured on the prints and the rate of translocation was determined by dividing the distance of migration by the time. The first numbers
in parentheses equal the number of cells observed, and the second numbers equal the total
number of 5 min observations for all the cells observed at each concentration of collagen.
mesenchyme. An SEM examination of neural tubes treated with trypsin and
dissected free of somitic mesenchyme revealed no somitic tissue (Fig. 1). Other
evidence has already been presented by one of us (Davis, 1980, p. 20-21).
RESULTS
Form, orientation, and social relations of migrating neural crest cells
Using the collagen fibrils of the HCL as a substratum, neural crest cells begin
to emigrate within 2-4 h. During this process, the majority of the cells are in
contact with each other (Fig. 2). Only a few single cells have been observed.
Characteristically, groups of cells become organized into elongate streams that
extend far away from the neural tube. In general, after 36 h in culture, a stream
consists of about ten cells and extends approximately 150 to 200 fim from the
neural tube. Cells which compose the streams line up largely in single file; however, at the base of the stream, they may be three to four abreast. Each cell is
elongate and oriented along the axis of migration, with the most distal cells
generally the most elongate of all (Fig. 2).
Locomotion of neural crest cells
X
NCH
Fig. 1. (a) Scanning electron micrograph of a neural tube isolated from a stage-14
chick embryo (see Methods) placed in 2% glutaraldehyde approximately 15min
after isolation. The tube is clean, with only the notochord (NCH) and neural crest
remaining attached, to the ventral and dorsal surfaces, respectively.
(b) An enlargement of the part of the tube included in the square of Figure 1 a.
In this region, the crest cells (NC) have already begun to move down the surface of
the neural tube. Neural crest cells appear highly spread, in close apposition to one
another, and have few blebs on their surfaces. In contrast, cells of the neural tube are
covered with many blebs. Occasionally, some cells of the overlying ectoderm (EC)
remain attached to the isolated tube.
34
E. M. DAVIS AND J. P. T R I N K A U S
Fig. 2. A frame from a time-lapse film of neural crest cells emigrating as streams of
cells from a neural tube (at the bottom and out of view) suspended in an HCL
containing 50/^g/ml of collagen, after 18 h in culture. Many cells are out of focus
because of the three-dimensional character of the preparation. Nomarski interference optics, x 230.
Obviously, as neural crest cells move in an HCL, they frequently adhere to
other cells. The stability of these adhesions depends in part on cell locomotor
behaviour. When cells in contact are moving in the same direction, they stay
together, suggesting either that their adhesions are quite stable or that their
identical direction of movement does not subject their adhesions to enough
stress to cause them to rupture. When two adhering cells move in opposite
directions, however, their adhesions rupture and the cells separate. Separation
also occurs when one cell of a moving adhering pair stops while the other
continues to move. In either case, before the cells separate they become more
elongate in the direction of movement and the collagen fibrils about them
become reoriented (Fig. 3). Note how one of the cells in Fig. 3 rounds up upon
separation and begins to bleb, as often occurs when cells de-adhere and retract
(Harris, 1973; Chen, 1981). When more than two cells are in contact and begin
to separate, neither the morphological distortion nor the rate of separation
appears as great.
Dependence of cell shape and rate of movement on the density of the HCL
Although emigration is not inhibited, there are distinct morphological differences between cells in HCL cultures containing low concentrations of
collagen (26 /*g/ml to 130 /*g/ml) and cells in cultures with higher concentrations
(200 /*g/ml to 390 /tg/ml) (Figs. 4, 5). In low concentrations, cells are usually
long and bipolar in shape (see Davis, 1980), whereas in high concentrations, cells
are more rounded. The average ratio of the length to width of cells was used to
15
8
10
6
7
14
80
105
130
200
260
390
61-120
(40)
0-76 ±0-49
(57)
0-78 ±0-48
(62)
0-70 ±0-43
(40)
0-84 ±0-54
(37)
0-39 ±0-24
(27)
0-32±0-18
(47)
0-44 ±0-26
0-60
(42)
0-63 ±0-35
(43)
0-74 ±0-50
(60)
0-94 ±0-56
(27)
O-86±O-51
(44)
0-38 ±0-26
(27)
0-34 ±0-22
(47)
0-44 ±0-25
181-240
(47)
0-68 ±0-44
(71)
0-75 ±0-47
(60)
0-76 ±0-57
(55)
0-79 ±0-46
(40)
0-41 ±0-26
(27)
0-48 ±0-29
(51)
0-40 ±0-32
121-180
(42)
0-64 ±0-59
(62)
0-83 ±0-60
(61)
0-78 ±0-50
(43)
0-87 ±0-54
(52)
0-37 ±0-20
(24)
O-3O±O16
(57)
0-31 ±018
(42)
0-75 ±0-48
(65)
0-66 ±0-41
(62)
0-74 ±0-51
(51)
0-74 ±0-40
(40)
0-46 ±0-34
(30)
0-41 ±0-35
(50)
0-35 ±0-20
241-300
Rates of translocation for each 60 min of filming
(ywm/min)
(32)
0-71 ±0-55
(47)
0-81 ±0-58
(32)
0-58 ±0-38
(32)
1 03 ±0-74
(37)
0-43 ±0-40
(21)
0-36±018
(43)
0-41 ±0-30
301-350*
(1-40)
018
(0-48)
0-63
(2-70)
001
(019)
0-85
(018)
0-86
(0-32)
0-74
(016)
0-99
120
(0-11)
0-91
(0-89)
0-37
(100)
0-31
(009)
0-93
(0-22)
0-83
(0-68)
0-50
(3-10)
0003
180
0-55)
0-60
(016)
0-87
(0-69)
0-49
(0-66)
0-51
(0-37)
0-70
(0-20)
005
(0-74)
0-46
240
(1-30)
0-20
(0-97)
0-33
(0-45)
0-65
(100)
0-30
(110)
0-26
(0-94)
0-35
(210)
004
300
350
(1-76)
0-45
(0-72)
0-42
(130)
0-21
(100)
0-32
(055)
0-57
(0-42)
0-68
(057)
0-56
Difference between mean rates of
translocation (/) P
* This is only 350 min, because the film inadvertently ran out at 350 rather than 360 min.
The average rate of movement of neural crest cells was determined from prints of the film at 10 min intervals. Consecutive 10 min observations
were grouped into 1 h periods, and the average rate of translocation of the cells for each 60 min period was determined. The numbers in parentheses
equal the total number of 10 min observations for all of the cells observed during each 60 min period. Note that the average rates of translocation
are smaller than those in Table 1. The reduced rates result from increasing the time interval between observations to 10 min, which incorporates
more time when cells are not moving.
The difference between the mean rates of translocation during each 1 h period was tested against the mean rate of translocation, determined for
the first hour of filming, using Student's Mest and assuming unknown but equal variances (in all cases, the F-ratio of the variances had a level of
significance greater than 005). The numbers in parentheses indicate the value of t for each test and the numbers without parentheses directly below
represent iheiVvalue^ l a practically all cases* the differences between means were statistically insignificant, with a level of significance greater than
005.
10
50
Collagen Number
cone.
of
cells
Table 2. The average rate of translocation of neural crest cells with increasing time in culture
§
I
1
36
E. M. DAVIS AND J. P. TRINKAUS
3^
FIGURE 3
A sequence from a time-lapse film of two neural crest cells emigrating from a neural
tube (out of view at the left) suspended in an HCL culture containing 50 /*g/ml of
collagen, after 20 h in culture. This sequence demonstrates the recoil of collagen
fibrils and the fast rate of separation when two cells break with one another. Times
are in minutes and seconds.
0:00 As the two cells begin to separate, three collagen fibrils can be seen. The parts
of fibrils a and b which are visible in the plane of focus are: 20 fim and 24 fim
long, respectively. Fibril c was not measured. The length of the cell at the left is
approximately 61 fim and that of the cell at the right approximately 37 fim.
2:38 The apparent lengths of fibrils a, and b and of the cell at the left have not
changed; however, the length of the cell at the right has increased to approximately
49 fim.
2:45 Within seconds, the cells separate and the cell at the right retracts approximately 14 fim to a length of about 23 fim and rounds up considerably. The cell at the
left has retracted little but is clearly less taut. The apparent length of fibril a
Locomotion of neural crest cells
37
I'lll
Fig. 4. A frame from a time-lapse film of neural crest cells emigrating from the
neural tube (lower left) in a low density hydrated collagen lattice, containing
26 /*g/ml of collagen, after 21 h in culture, x 480.
Fig. 5. A frame from a time-lapse film of neural crest cells emigrating from the
neural tube (lower left) in a high density hydrated collagen lattice, containing
394 /*g/ml of collagen, after 21 h in culture, x 480.
quantify the differences in their morphology (Table 1). Increasing the concentration of collagen in HCL cultures to 200 /*g/ml or more causes a two-fold reduction in the length-to-width ratio of moving neural crest cells.
The rates of translocation of neural crest cells in different concentrations of
collagen are shown in Table 1. The average rate of translocation was found to
be clearly dependent on the collagen concentration. In cultures containing low
concentrations, the average rate was approximately 1 /on/min, whereas at
higher concentrations, it was reduced to approximately 0-5 /tm/min.
increases to about 24/*m, as the cell at the left moves slightly to the left. Fibril b
decreases in length to about 22 fim, when the protrusion of the cell at the left loses
contact with the cell at the right, shortens, and retracts toward the left. Fibril c,
previously slanted towards the right (times 0:00-2:38), now has moved nearly 6 /fm
to the left.
2:53 The visible length of fibril a has now increased to about 27 fim, while fibril b
apparently shortens and becomes curved (distance between reference points 18 /«n).
The cell at the left has clearly retracted and its length has decreased to approximately 57 fim.
3:45 The apparent length of fibril a is now about 29 fim and fibril b has becotne
curved, with the distance between its two reference points decreasing to 16 fim. The
cell at the left retracted an additional 8 fim and its length has decreased to approximately 49 fim. The cell at the right is now fully retracted and is blebbing actively.
x480.
38
E. M. DAVIS AND J. P. TRINKAUS
Table 3, Average rate of tramlocation of neural crest cells at
various distances from the neural tube
Distance
from
HCL cultures
containing 50-105 /*g collagen/ml
average rate of translocation
HCL cultures
containing 200-360 fig collagen/ml
average rate of translocation
A
A
(/im)
/tm/min
N
0-20
20-40
40-60
60-80
80-100
100-120
120-140
140-160
160-180
l-10±0-82
1-00 ±0-60
1-11 ±0-70
0-96±0-63
0-98 ±0-70
1-22 + 0-88
MO±O-83
l-32±0-88
l-32±0-88
199
153
196
162
42
23
16
4
P
4
—
—
0-30
1-18
0-69
0-42
1-25
0-32
0-80
0-80
0-76
0-24
0-49
0-67
0-21
0-75
0-43
0-43
r
\
/*m/min
0-53 ±0-30
O-53±O-36
0-54±0-37
0-56±0-50
N
t
P
338 —
—
168 017 0-86
94 0-35 0-73
45 0-60 0-55
Data from HCL cultures containing 50, 80, and 105 /tg of collagen/ml were grouped
together for this analysis since there was no statistical difference in the rates of translocation
in these cultures. These rates are based on observations of 33 cells.
N equals the number of measurements made at each segment of 20 fim. Since few neural
crest cells emigrate long distances from the neural tube under these conditions, the number
of observations decreases as the distance from the neural tube increases after 80 /im.
The differences between the mean rates of translocation at various distances from the
neural tube were tested with the /-test, assuming unknown but equal variances, and the level
of significance of the test is given by the P-value. In all cases, the differences between the
means are insignificant.
Data from the HCL cultures containing 200-360/tg of collagen/ml were also grouped
together for this analysis since there was no statistical difference in the rate of translocation
in these cultures. These rates are based on observations of 27 cells.
Possible variation in the rate of translocation with time in culture and distance
from the neural tube
To test for variations in the rate of translocation with time and distance,
estimates of the rate of translocation were made at 60 min intervals from the
beginning to the end of filming, i.e. during a period of nearly 6 h. The differences
between the rates during the first hour and the succeeding hours were then
compared. They were found to be statistically insignificant; no correlation was
observed between the average rates of translocation and the duration of migration (Table 2). Thus, estimates of the average rate of translocation of neural
crest cells are not significantly biased by the time in culture at which the estimates
are made.
Variations in the rate of translocation with increasing distance from the
neural tube could be significant in HCL cultures containing low concentrations
of collagen, since the average rate of translocation and the distance the cells
travel are both great. To test for this possibility, estimates of the rate of trans-
Locomotion of neural crest cells
39
B
10/im
Fig. 6. The course traversed by neural crest cells emigrating from neural tubes su$pended in hydrated collagen lattices was plotted from time-lapse films. The initial
position of each cell, the course of its migration, and changes in its direction Of
translocation are designated by solid lines and arrows, respectively. The curvaceous
line on the left side of each tracing represents the edge of the neural tube. In both
low and high density hydrated collagen lattice cultures, neural crest cell locomotion
from the neural tube appears to be directional. Their paths of migration leads almost
always away from the neural tubes. (A) The neural tube was suspended in a lowdensity hydrated collagen lattice containing 26 /*g/ml of collagen. The illustrated
sequence represents an elapsed time of 100 min. (B) The neural tube was suspended
in a high-density hydrated-collagen lattice containing 394/ig/ml of collagen. The
illustrated sequence represents an elapsed time of 500 min.
location made at 20 /tm intervals of distance from the neural tube were compared. The difference between the rates of translocation at various distances
from the neural tube were found to be statistically insignificant; no correlation
was observed between the average rate of translocation and the distance cells
have migrated from the neural tube (Table 3).
Directionality of migration of neural crest cells
Once neural crest cells are moving within an HCL, the paths of migration are
largely perpendicular to the neural tube (Fig. 6). Cells were seldom observed to
move to the left or right and never observed to move above or below the plane
of focus. Since the route of migration is so direct, the time spent moving away
from and towards the neural tube and the rate of cell movement must be the
main factors determining the distance they travel. Analysis revealed that they
move faster and spend more time moving away from the neural tube than
40
E. M. DAVIS AND J. P. TRINKAUS
Locomotion of neural crest cells
41
Table 4. Relative amount of time spent by neural crest cells moving
away from and towards the neural tube
/o
Collagen cone.
Q*g/ml)
50-105
200-390
,
of time of observation
N
Movement
from
neural tube
Movement
toward
neural tube
Immobile
1625
1312
58±7
68 ±12
3O±9
19±7
12±4
12±9
Prints offilmswere analysed at 5 min intervals to determine the percentage of the time of
observation (filming) the 33 cells in HCL cultures containing 50 to 105/ig/ml of collagen
and the 27 cells in HCL cultures containing 200 to 360 /*g/ml of collagen remain stationary
or migrate away from or towards the neural tube, //equals the total number of measuretnents
made.
toward it (at least in low concentrations of collagen). Statistically significant
differences were observed in both instances (Tables 4, 5).
The next question is that of directionality itself. What factors contribute to
directional movement of neural crest cells in an HCL? One could be the orientation of the collagen fibrils. If these fibrils were aligned perpendicular to the
neural tube, they might provide a kind of contact guidance and by reducing or
preventing sideward movements contribute to directionality. The fibrils were
FIGURE 7
Orientation of thefibrilsof the HCL with respect to the neural tube and the direction
of migration of emigrating neural crest cells. An isolated neural tube with adhering
neural crest was suspended in an HCL culture containing 50 /*g/ml of collagen a$d
cultured over night to permit time for the neural crest cells to migrate away from
the neural tube. Polarization microscopy was used to examine the orientation of the
collagen fibrils of the HCL (see Methods). Photographs of the positive (A', B', etc.)
and negative (A, B, etc.) retardation images were taken as the culture was rotated
0, 60, 90, 120 and 180°. (A, A') Retardation images of the collagen lattice before the
culture was rotated. The majority of the birefringent collagenfibrilsradiate from the
neural tube; however, upon close examination, some birefringent fibrils appear
oriented perpendicular to the longer fibres of the HCL. (B, B') The culture was
rotated 60°. Note that the intensity of the retardation images radiating from the
neural tube has decreased and that no new birefringent collagen fibrils appear.
(C, C ) The culture was rotated 90°. As in A, A', the birefringent images radiate frpm
the neural tube; however, their polarity is reversed. (D, D') The culture was rotated
120°. The intensity of the retardation images has decreased; however, the orientation of the collagen fibrils with respect to the neural tube remains unchanged.
(E, E') The culture was rotated 180°. The retardation images appear as in A, A'.
Since most of the birefringent images of the HCL radiate from the neural tube like
the spokes from the hub of a wheel, regardless of the degrees of rotation, it may be
concluded that much of the HCL is highly ordered. In addition, it is important to
point out that the orientation of the HCL is nearly identical to that of the path of
migration of emigrating neural crest cells.
42
E. M. DAVIS AND J. P. TRINKAUS
Table 5. Average rate of translocation of neural crest cells migrating
away from and toward the neural tube
Collagen
cone.
(/*g/ml)
Rate of movement
away from
neural tube
(/im/min)
Rate of movement
towards
neural tube
(/^m/min)
50-105
200-390
1-66 ±0-45 (656)
0-33 ±0-21 (696)
0-58 ±0-43 (366)
0-30 ±0-21 (375)
Difference between
Means
/
2-83
2-58
P
0005
001
Prints of films were analyzed at 5 min intervals to determine the average rate of translocation away from and toward the neural tube. Thirty-three cells were observed in cultures
containing 50 to 105 /ig/ml of collagen and 27 in cultures containing 200 to 390 /^g/ml of
collagen. Time of observation for each cell ranged from 200 to 360 minutes. The numbers in
parentheses indicate the total number of observations. The differences between the rates of
translocation were tested with Student's /-test, assuming unknown but equal population
vatiances. The P-value gives the level of significance.
examined with polarization microscopy to deteimine their orientation. As can
be seen in Fig. 7, the majority of the fibres were aligned perpendicular to the
neural tube. Since the cells are similarly aligned, this would appear to be a clear
case of contact guidance. However, since contact guidance can only provide
orientation, toward or away from the tube (see Trinkaus, 1976), some other
factor must be added to give directionality, i.e. movement away from the tube.
It is possible that this directionality could come from cell-to-cell interactions.
Effect of cell-to-cell contact on cell migration
One possibility is that the greater time spent by neural crest cells moving away
from the neural tube (Table 4) is related to cell-to-cell contact. Following the
procedure of Abercrombie & Heaysman (1966), we found no correlation
between the number of cell-to-cell contacts per cell and duration of emigration
(Table 6). These results suggest that the increased time neural crest cells spend
moving away from the neural tube is not influenced by cell-to-cell contacts.
Although the duration of directional migration is unaffected, cell contacts
might nevertheless influence the rate at which neural crest cells translocate
(Abercrombie & Heaysman, 1953; Martz, 1973), and this, in turn, could have
a significant effect on the distance cells migrate, since the chance of collisions
between cells would be greater for cells migrating toward the neural tube than
away from it. Analysis of the effect of cell-to-cell contact on rate of translocation,
however, revealed no correlation between these two variables (Table 7). Regardless of the contact number, the average rate of translocation remained unaltered.
Since the duration and rate of translocation of neural crest cells is greater
when moving away from the neural tube, cell-to-cell contacts might influence
cells differently, depending on their direction of movement. Such a difference
might be obscured by analyses that incorporated data from cellular movement
Locomotion of neural crest cells
43
Table 6. Correlation of contact number with the migratory
activity of neural crest cells
Average
% of time observed
Collagen cone. 50-105 fig/ml
number
EM
IM
Stationary
N
0
0-5
10
1-5
20
2-5
30
3-5
40
50
53
56
53
63
59
71
67
53
33
43
27
36
25
31
16
20
40
17
4
16
12
12
10
10
13
7
All
91
361
120
386
78
87
15
15
% of time observed
Collagen cone. 200-390 /*g/ml
A
EM
IM
Stationary
N
57
66
55
83
54
86
67
68
80
25
24
30
15
12
7
21
24
14
17
10
15
2
33
7
12
8
6
387
68
343
95
161
28
103
34
93
Prints of the films were made at 5 min intervals. The data were analyzed to determine if
a correlation exists between the average contact number and the duration of emigration (EM),
immigration (IM), and of the stationary state. Thirty-three cells were observed in 50 to
105 /tg/ml of collagen and 27 cells in 200-390 /*g/ml of collagen. Time of observation for
each cell ranged from 200 to 360 min. TV equals total number of observations. Average
contact number was determined by averaging the number of cells that appeared to be in
contact with the cell being analyzed during two consecutive 5 min intervals. No correlation
was observed between the percent of time in emigration and average contact number in
cultures containing 50 to 105/tg/ml of collagen or 200 to 390/tg/ml of collagen, when
tested using the coefficient of correlation (r), where r = 0-57, / = 1-85, degrees of freedom
(d.f.) = 7, and P = 011 for cultures with low concentrations of collagen and r =* 047,
/ = 14, d.f. = 7 and P = 020 for cultures with high concentrations.
in both directions (as in Table 7). To test for this possibility, the rate of translocation, as related to contact number, was determined separately for cells
moving in only one direction. It is clear from the results, summarised in
Tables 8 and 9, that no coirelation is evident; regardless of the contact number,
the rate of directional translocation remained unaltered. Thus, the net directional
displacement of neural crest cells away from the neural tube in an HCL is not
influenced by cell-to-cell contacts and, in consequence, other factors must
influence their directional migration.
Changes in cell-to-cell contact and their effect on the rate of translocation
It was noted above (p. 34) that the rate of separation between two cells losing
contact with each other appeared to depend in part on the number of remaining
adhesions with neighbouring cells. To test this impression, the average rate of
translocation for cells (1) making, (2) breaking, or (3) maintaining a constant
contact number was determined, as described by Abercrombie & Heaysman
(1954). Only cells losing contacts had a statistically greater rate of movement
(Table 10). Since only cells losing contacts translocate at a significantly greater
rate, the effect of cell-to-cell contact on this form of movement was determined.
44
E. M. DAVIS AND J. P. TRINKAUS
Table 7. The effect of cell-to-cell contact on the average rate of
translocation of neural crest cells
Collagen cone. 200-390/ig/ml
Collagen cone. 50-105 /ig/ml
Difference
between means
Rate of
translocation
Average
Difference
between means
Rate of
translocation
A
A
1
number
0
0-5
10
1-5
20
2-5
30
3-5
40
/im/min
/
N
—
0-74±0-53 258
0-92 ±0-62 60
0-67 ±0-44 211
0-80 ±0-49 72
0-70 ±0-48 230
0-83 ±0-55 42
0-81 ±0-59 88
8
0-62 + 0-56
7
0-72 ±0-44
/fm/min
p
2-28
1-92
0-84
0-83
1-31
0-61
0-62
002(316)
013 (467)
0-40 (328)
0-38 (486)
0-19 (298)
0-54 (344)
0-52 (264)
011 0-96 (263)
N
0-40 ±0-26 177
0-42 ±0-23 31
0-37 ±0-26 179
0-46 ±0-33 28
0-40 ±0-23 125
0-48 ±0-44 20
0-41 ±0-24 51
0-39±0-19 26
0-40 ±0-25 55
t
P
—
—
0-62 0-54 (206)
0-75 0-44 (354)
119 0-23 (203)
0-51 0-60 (300)
1-32 018 (195)
0-45 0-65 (226)
003 0-97 (201)
0-24 0-81 (230)
Prints of the films were made at 5 min intervals and the rate of translocation and average
contact number (i.e. the number of cells that appear to be in contact with the cell being
analyzed) were determined for two consecutively paired 5 min intervals. Thirty-three cells
were analyzed in cultures with 50 to 105 /*g/ml of collagen and 27 cells in cultures with 200
to 390 /*g/ml of collagen. Time of observation for each cell ranged from 200 to 360 min.
Nequals the total number of paired 5 min intervals. The differences between the mean rates
of translocation for cells with an average contact number of 0-5 or greater were tested against
the mean rate of translocation of cells with zero cell contacts using Student's /-test and
assuming unknown but equal variances. (In all cases the F-ratio of variances had a level of
significance greater than 0-05). The level of significance for the /-test is given by the P-value.
Numbers in parentheses equal the degrees of freedom. (See also Tables 8 and 9.)
Table 8. Effect of contact number on the rate of translocation of
neural crest cells moving away from the neural tube
Collagen cone. 5O-15O/*g/ml
Collagen cone. 200-390 /tg/ml
A
A
i
Average
number
0
0-5
10
1-5
20
2-5
30
3-5
40
Rate of
translocation
Difference
between rates
A
A
Rate of
translocation
Difference
between rates
A
K
r
r
ftm/m'm
N
0-67 ±0-44 163
0-67 ±0-48 39
0-68 ±0-46 145
0-64 ±0-44 49
0-67 ±0-48 174
0-70 ±0-41 36
0-55 ±0-36 46
4
0-80 ±0-65
—
—
/
P
—
0-28 0-78 (200)
019 0-85 (306)
0-41 0-68 (210)
0 0 2 0-98 (355)
0-40 0-69 (197)
0-98 0-36 (207)
0-60 0-54 (165)
—
—
/Mn/min
N
0-33 ±0-20 120
0-42 ±0-21 25
0-32 ±0-22 138
0-45 ±0-34 22
030 ±0-20 130
0-34 ±0-20 18
0-35 ±0-22 53
O-35±O16 23
0-35 ±0-21 47
See the legend to Table 7.
/
P
—
1-84
0-36
1-52
1-39
0-25
0-57
0-38
0-58
007 (143)
0-72 (256)
003 (24)
0-16(248)
0-98 (136)
0-56(171)
0-70 (141)
0-56 (165)
Locomotion
45
of neural crest cells
Table 9. Effect of contact number on the rate of translocation of
neural crest cells moving toward the neural tube
Collagen cone.
5 0 - 1 0 5 /*g/ml
Collagen cone . 200-390/tg/rql
A
Rate oiF
translocation
A
Difference
between rates
A
number
0
0-5
10
1-5
20
2-5
30
3-5
40
/*m/min
N
0-61 ±0-46 116
0-77 ±0-57 24
0-55 ±0-34 88
0-63 ±0-52 36
0-49 ±0-33 74
8
0-44±0-15
0-66 ±0-58 19
—
—
—
—
t
1-42
104
-018
204
110
-0-41
—
—
0-15 (138)
0-29(202)
0-85(150)
004(188)
0-27(122)
0-68(133)
—
—
/*m/min
0-31 ±0-25
O-36±O-22
0-28 ±0-20
0-39 + 0-24
0-27±0-15
0-22 ±012
0-38 + 0-28
0-22 ±012
0-21 ± 0 1 1
Difference
between rates
A
A
P
(
Rate oir
translocation
t
1
N
85
10
72
10
64
6
16
6
14
•»
t
P
—
0-62
0-92
0-97
119
0-87
0-89
0-87
1-47
0-53 (93)
0-36(155)
0-34 (93)
0-24 (140)
0-38 (89)
0-37 (99)
0-38 (89)
014 (97)
_
See the legend to Table 7.
Table 10. Changes in contact number and the associated rate of translocation
Collagen cone . 50-105 fig/ml
Changes
in contact
number
0
+1
-1
Rate of
translocation
[im/mxn
N
0-99 ±0-63 1092
106 ±0-68 159
l-26±0-91 161
Differences
between rates
/
P
Collagen cone. 200-390 /*g/ml
Rate of
translocation
/tm/min
N
0-51 ±0-30 932
1-24 0-20(1249) 0-60±0-52 72
-9-57 0001 (241) 0-62 ±0-39 104
Differences
between rates
/
P
1-45 015 (1004)
2-83 0005 (1034)
Prints were made at 5 min intervals. Thirty-three cells were analyzed in HCL cultures
containing 50 to 105/tg/ml of collagen and 27 cells in HCL cultures containing 200 to
390 /tg/ml of collagen. The time of observation ranged from 200 to 360 min. N equals the
total number of observations. The data were segregated into three groups: cells that did not
change contact number (0): cells that increased their contact number by one (+1): and cells
that lost one cell-to-cell contact (-1). The difference between the mean rates of translocation
of cells changing contact number was tested against the mean rate of translocation of cells
maintaining a constant number, using Student's /-test and assuming unknown but squal
variances for cells making an additional contact in HCL cultures containing 50 to 105 /*g/ml
of collagen and unknown and unequal variances for the remaining /-test. The differences
between the variances were determined by the F-ratio of the variances with a level of Significance of 005. The significance of the difference between the mean rates of translocation is
given by the P-value. The number in parentheses equals the degrees of freedom.
46
E.M.DAVIS AND J. P. TRINKAUS
Table 11. Contact number and the rate of separation of
cells losing mutual contacts
Collagen cone. 50-105 /*g/ml
Collagen cone. 200-390 /*g/ml
Rate of
separation
Rate of
separation
Contact
number
A
,
/*m/min
N
0
1
2
3
1-42 ±0-94
M0±0-87
M5±0-70
0-76 ±0-40
64
74
36
10
Differences
between rates
>
A
,
t
*
,
P
2•09 004 (136)
1•52 013 (98)
2•21 003 (72)
K
*
/*m/min
N
0-74 ±0-57
0-69 ±0-38
0-47 ±019
0-48 ±0-23
26
32
13
17
Differences
between rates
,
*
t
P
—
0-42 0-67 (56)
3-53 0001 (37)
3-31 0002 (41)
Prints of the films were made at 5 min intervals. The rate of translocation and average
contact number (i.e., the number of cells that appear to be in contact with the cell being
analyzed) for cells losing cell-to-cell contacts was determined at each 5 min interval. Thirtythree cells were analyzed in HCL cultures containing 50 to 105 /*g/ml of collagen and 27 cells
in HCL cultures containing 200 to 390 /^g/ml of collagen. The time of observation ranged
from 200 to 360 minutes. N equals the total number of 5-min observations. The difference
between the mean rates of separation for cells with one or more contacts was tested against
the mean rate of separation of cells with a contact number of zero, using Student's /-test and
assuming unknown but equal variances. The level of significance of the /-test is given by the
P-value and the number in parentheses equals the degrees of freedom.
To determine if a correlation exists between contact number and the rate of separation of
cells losing cell-to-cell contacts, the coefficient of correlation (r) was calculated and tested to
determine if it was greater than zero. Though the value of r was greater than 0-90 for both
cultures containing 50 to 105/*g/ml (r = 0-92, n = 4, / = 3-32, and P = 008) and 200 to
390/ig/ml (r = 0-91, n = 4, / = 3-13, and P = 009) of collagen, r was not significantly
greater than zero at a level of significance of 005.
The correlation between the rate of separation of two cells losing mutual
contacts and the remaining number of cell-to-cell contacts retained after separation was equivocal. The average rate of separation tends to decrease as contact
number increases; however, no statistically significant correlation was observed
between contact number and the rate of separation of cells losing contact
(Table 11).
DISCUSSION
This study has explored some of the factors that might influence the migratory
behaviour of chick neural crest cells. Since the movement of these cells cannot
be observed directly within the embryo, neural tubes with intact neural crest
were suspended in a transparent hydrated collagen lattice in vitro (Elsdale &
Bard, 1972). This culture system was chosen not only because it permits direct
observation of cell movement, but because it resembles, in part, the threedimensional fibrillar matrix through which neural crest cells migrate normally
in vivo (see Noden, 1978, for review). It should be emphasized, however, that
although such an HCL consists primarily of collagen, a normal constituent of
the matrix, it lacks other normal constituents, including glycosaminoglycans.
Locomotion of neural crest cells
47
Neural crest cells were found to translocate in a highly directional manner in
an HCL, spending 74 % of their time, on the average, moving away from the
neural tube. Analysis of the effect of cell-to-cell contacts on cell movement
revealed no correlation between the number of contacts between cells and their
rate or direction of movement. Since neural crest cells are perfectly capable of
forming firm adhesions with neighbouring cells under these conditions, as
shown by their physical distortion when two adhering cells move apart, it seems
unlikely that this could be attributed to weak intercellular adhesions. In fact, it
has been shown (Johnston & Listgarten, 1972; Ebendal, 1977) that neural crest
cells form cell-to-cell contacts in vivo that are similar to those formed in vitro on
a planar substratum by chick heart fibroblasts (Heaysman & Pegrum, 1973),
whose rate and direction of translocation is clearly modulated by cell-to-cell
contacts. It seems clear from these observations that neural crest cells moving
within an HCL do not show contact inhibition of movement, as originally
defined by Abercrombie & Heaysman (1953, 1954). Thus, other factors must
be responsible for the directional movement of these cells under these conditions.
It goes without saying that these results do not tell us whether the striking
directional migration of neural crest cells as they leave the neural tube in Situ
during normal development is independent of contact inhibition or not. The
contact behaviour of cells in vitro does not necessarily reflect their normal
behaviour in vivo (see, for example, Lesseps, Lapeyre & Hall, 1979). These
results do suggest, however, that contact inhibition may not operate for these
cells in situ within the embryo and they compel us to pay more attention to
other possible causes of the directionality of neural crest cells (and of other
directionally moving cells as well).
One possibility is that directional migration is determined intrinsically for
each cell; i.e. once cells begin to move in a certain direction they will tend to
persist in that direction. This is not implausible. Two different forms of fibroblasts, primary cultures of chick heart fibroblasts (Abercrombie & Gitlin, 1964;
Abercrombie & Heaysman, 1966) and 3T3 cells (Gail & Boone, 1970), have
been observed to continue their directional, non-random translocation in vitro
for long periods of time, ranging from 2-5 to 5 h, in the absence of cell contact
and orientation of the substratum. Clusters of fish melanoma cells also show
such persistence in vitro (Kolega, 1981). If neural crest cells in an HCL similarly
maintained their initial direction of movement, it would be away from the neural
crest, exactly what we have observed. Retraction induced spreading (Chen,
1979; Dunn, 1980), whereby the protrusive activity of the advancing leading
edge of a fibroblast is augmented by the retraction of its trailing edge, could be
the mechanism (Trinkaus, 1980). However, our observations of neural crest Cells
in an HCL offer no support for this hypothesis. No spurt of protrusive activity
at the leading edge was observed when the trailing edge of these cells retracted
(Davis, 1980). This does not eliminate persistence as a possible explanation of
48
E. M. DAVIS AND J. P. TRINKAUS
neural crest directionality, but is does suggest that if it occurs it is by some other
mechanism.
The other possibility, of course, is that factors in the cellular environment
(other than contacts with other cells) have a strong directional influence on the
migration of neural crest cells. The experiments of Weston & Butler (1966), Le
Douarin & Teillet (1974) and Noden (1975) all have provided evidence that
transplanted neural crest cells tend to follow the pathways normal for that level
of the neural tube; i.e. the direction of their migration is strongly influenced by
the tissue environment.
In the present study, one element of the environment - the collagen fibrils appears to have a strong influence on the migratory behaviour of the cells.
Polymerization of these fibrils apparently begins soon after a neural tube is
placed into the solution of collagen, and they soon become aligned perpendicular
to the neural tube, radiating from it like spokes of a wheel. The cause of this
alignment is unknown, but it seems probable that the neural crest cells themselves are in part responsible, as they pull and tug on these fibrils during their
migration (see Harris, Wild & Stopak, 1980; Stopak & Harris, 1980). Markwald,
Fitzharris, Bolender & Bernanke (1979) have shown, for example, that microfibrils of the cell-free cardiac jelly matrix of the chick embryo are unoriented
prior to cellular invasion. But, with entrance of the cushion cells, those fibrils
that are associated with motile cellular extensions become oriented. As the
authors point out, this association suggests but does 'not prove' that these
oriented microfibrils constitute the natural in vivo substratum of the cell. However the collagen fibrils become oriented in an HCL in our experiments, neural
crest cells quickly become oriented and aligned with them in what appears to be
typical contact guidance. It is not known how the cells align in this manner. Of
course, the simplest explanation, and perhaps the most probable, would be that
the aligned collagen fibrils physically restrain the cells from moving tangentially.
In any case, as already emphasized (see p. 29), such an orienting influence
cannot in and of itself give directionality (see Trinkaus, 1976). Further, a crucial
element in this demonstration of possible contact guidance is unfortunately
lacking. We have not been able to construct an HCL containing a neural tube
in which the collagen fibrils are unoriented. Until such a control can be provided,
these remarks can only be regarded as speculative.
Since we chose an HCL for these studies in order to try to simulate, though
crudely, certain aspects of the normal environment within the embryo, it is of
interest that the fibrillar matrix through which neural crest cells normally
migrate in the embryo is likewise highly aligned, at least in amphibians (Lofberg,
1976; Lofberg & Ahlfors, 1978). It is not clear whether this is so for the avian
embryo. Ebendal (1977) has observed some fibre alignment near the neural crest
with SEM, but Bancroft & Bellairs (1976) and Tosney (1978) have not. Tosney
(1978) points out that this failure to observe alignment of the fibrillar extracellular matrix may be due in part to fixation problems. She may well be correct;
Locomotion of neural crest cells
^9
for an HCL fixed with glutaraldehyde collapses upon critical-point drying,
forming a flat mat, in which both the orientation and three-dimensional distribution of the collagen fibres is destroyed (Davis, unpublished results).
Another environment factor that has an important influence on the locomotor
behaviour of neural crest cells in an HCL is the concentration of the collagen.
Although it does not modify the direction of cell movement, it has important
quantitative effects on cell shape, rate of translocation, distance of migration,
and the rate of separation of two adhering cells. These responses are not understood. Perhaps cells move more readily in lower concentrations because these
provide sufficient substrata for movement but little impediment, whereas higher
concentrations are so dense that they physically impede movement.
Mutual negative chemotaxis might also be a factor in this directional migration, as for amphibian melanoblasts, which are a derivative of the neural crest
(Twitty, 1944; Twitty & Niu, 1954). This seems unlikely, however, since neural
crest cells in HCL cultures tend to remain in contact with one another during
migration. Moreover, continuous perfusion of growth medium through these
cultures does not alter the directionality of the cells (Davis, unpublished
results).
An important feature of neural crest cells within an HCL is their tendency to
move in streams, the leading edge of one cell adhering to the trailing edge of
another cell. This is no doubt an inevitable consequence of their moving in the
same direction. Not surprisingly, this same phenomenon has been observed
elsewhere, both in vivo (Bancroft & Bellairs, 1976) and in vitro. Chick heart
fibroblasts, for example, form similar cell streams occasionally in vitro (Ambrose,
1961), albeit, unlike neural crest cells in an HCL, they show contact inhibition
of movement. Apparently, leading-to-trailing edge collisions do not elicit
contact inhibiting responses from either neural crest or chick heart fibroblasts
in vitro.
What then is the mechanism of the directional migration of neural crest cells,
either in an HCL in vitro or normally in vivol There is no evidence for contact
inhibition, once they have left the crest. By elimination, it seems possible that
the intrinsic resistance of the cells themselves to changes in direction is a primary
force in their directional movement. They no doubt move away from the crest
initially, because there is no other direction available to them. What causes them
to persist in this path of movement is the question. All we can say at present is
that retraction-induced protrusive activity appears not to be at work and that
contact guidance may prevent them from moving sideways.
We are indebted to Raymond E. Keller and Wen-Tien Chen for discussion and assistance.
This research has been supported by a grant from the NIH (CA22451) to J.P.T. and by
postdoctoral fellowships from the American Cancer Society (PF-1330) and a Cell and
Developmental Biology Training Grant (USPHS-HD-OOO32-13) to the Department of
Biology of Yale University to E.M.D.
50
E. M. DAVIS AND J. P. T R I N K A U S
REFERENCES
M. & GITLIN, G. (1964). The locomotory behaviour of small groups of fibroblasts. Proc. Royal Soc. 162, 289-302.
ABERCROMBIE, M. & HEAYSMAN, J. E. M. (1953). Observations on the social behaviour of
cells in tissue culture. I. Speed of movement of chick heart fibroblasts in relation to their
mutual contacts. Expl Cell Res. 5, 111-131.
ABERCROMBIE, M. & HEAYSMAN, J. E. M. (1954). Observations on the social behavior of cells
in tissue culture. II. Monolayering of fibroblasts. Expl Cell Res. 6, 293-306.
ABERCROMBIE, M. & HEAYSMAN, J. E. M. (1966). The directional movement of fibroblasts
emigrating from cultured explants. Ann. Med. Exp. Fenn. 44, 161-165.
AMBROSE, E. J. (1961). The movements offibrocytes.Expl Cell Res. 8, 54-73.
ANDERSON, T. F. (1951). Techniques for the preservation of three-dimensional structures in
preparing specimens for the electron microscope. Trans. N.Y. Acad. Sci. 13, 130-133.
BANCROFT, M. & BELLAIRS, R. (1976). The neural crest cells of the trunk region of the chick
embryo studied by SEM and TEM. Zoon 4, 73-85.
BARD, J. B. L. & HAY, E. D. (1975). The behavior of fibroblasts from the developing avian
cornea. /. Cell Biol. 67, 400-418.
CHEN, W.-T. (1979). Induction of spreading during fibroblast movement. / . Cell Biol. 81,
684-691.
CHEN, W.-T. (1981). Surface changes during retraction-induced spreading of fibroblasts in
culture. J. Cell Sci. (in the Press).
DAVIS, E. M. (1980). Translocation of neural crest cells within a hydrated collagen lattice.
/. Embryol. exp. Morph. 55, 17-31.
DUNN, G. A. (1980). Mechanisms of fibroblast locomotion. In Cell Adhesion and Motility
(ed. A. S. G. Curtis & J. D. Pitts), pp. 409-423. London: Cambridge University Press.
EBENDAL, T. (1977). Extracellular matrix fibrils and cell contacts in the chick embryo:
Possible roles in orientation of cell migration and axon extension. Cell Tissue Res. 175,
439-458.
ELSDALE, T. & BARD, J. (1972). Collagen substrata for studies on cell behavior. /. Cell Biol.
54, 626-637.
GAIL, M. H. & BOONE, C. W. (1970). The locomotion of mouse fibroblasts in tissue culture.
Biophys. J. 10, 980-993.
HAMBURGER, V. & HAMILTON, H. (1951). A series of normal stages in the development of the
chick embryo. /. Morph. 88, 49-92.
HARRIS, A. K. (1973). Cell surface movements related to cell locomotion. In Locomotion of
Tissue Cells, Ciba Foundation Symposium 14 (new series), pp. 2-20. Amsterdam: Elsevier,
North Holland.
HARRIS, A. K., WILD, P. & STOPAK, D. (1980). Silicone rubber substrata: A new wrinkle in
the study of cell locomotion. Science 208, 177-179.
HEAYSMAN, J. E. M. & PEGRUM, S. M. (1973). Early contacts between fibroblasts: An ultrastructural study. Expl Cell Res. 78, 71-78.
HORSTADIUS, S. (1950). The Neural Crest. London, New York: Oxford University Press.
JOHNSTON, M. C. & LISTGARTEN, M. A. (1972). Observations on the migration, interaction,
and early differentiation of orofacial tissue. In Developmental Aspects of Oral Biology (ed.
H. C. Salukin & L. A. Bavetta). New York: Academic Press.
KOLEGA, J. (1981). The movement of cell clusters in vitro: morphology and directionality.
/. Cell Sci. (in the Press).
LE DOUARIN, N. M. & TEILLET, M. A. (1974). Experimental analysis of the migration and
differentiation of neuroblasts of the autonomic nervous system and of neuroectodermal
mesenchymal derivatives, using a biological cell marking technique. Devi Biol. 41,
162-184.
LESSEPS, R. J., LAPEYRE, M. V. & HASL, M. V. (1979). Tissue culture evidence on the control of the switch from contact inhibition of cell movement to overlapping behaviour in
annual fish embryos of Nothobranchius korthausae. J. exp. Zool. 210, 521-528.
ABERCROMBIE,
Locomotion of neural crest cells
$1
J. (1976). Scanning and transmission electron microscopy of early neural crest
migration and extracellular fiber systems of the amphibian embryo. / . Ultrastruct. Res.
54, 484a.
LOFBERG, J. & AHLFORS, K. (1978). Extracellular matrix organization and early neural crest
cell migration in the axolotl embryo. Zoon 6, 87-101.
MARKWALD, R. R., FITZHARRIS, T. P., BOLENDER, D. L. & BERNANKE, D. H. (1979). Structural analysis of cell: matrix association during morphogenesis of atrioventricular cushion
tissue. Devi BioL 69, 634-654.
MARTZ, E. (1973). Contact inhibition of speed in 3T3 and its independence from postconfluence inhibition of cell division. /. cell Physiol. 8, 39-48.
NODEN, D. M. (1975). An analysis of the migratory behavior of avian cephalic neural crest
cells. Devi BioL 42, 106-130.
NODEN, D. M. (1978). Interaction directing the migration and cytodifferentiation of avian
neural crest cells. In The Specificity ofEmbryological Interactions (ed. D. Garrod). Chaprtfan
& Hall.
STOPAK, D. & HARRIS, A. K. (1980). Structuring of collagen matrices by cell traction. / . Cell
Biol.%1, 117a.
TOSNEY, K. W. (1978). The early migration of neural crest cells in the trunk region of the
avian embryo: An electron microscopic study. Devi BioL 62, 317-333.
TRINKAUS, J. P. (1976). On the mechanism of metazoan cell movements. In The Cell Surface
in Animal Embryogenesis and Development (ed. G. Poste & G. L. Nicolson). Cell Surface
Rev. 1, 225-329.
TRINKAUS, J. P. (1980). Formation of protrusions of the cell surface during tissue cell movement. In 'Tumor Cell Surfaces and Malignancy' (ed. R. Hynes & G. Fox), Progress in
Clinical & BioL Res., 41, 887-906.
TWITTY, V. C. (1944). Chromatophore migration as a response to mutual influences of the
developing pigment cells. J. exp. Zool. 95, 259-290.
TWITTY, V. C. & Niu, M. C. (1954). The motivation of cell migration, studied by isolation
of embryonic pigment cells singly and in small groups in vitro. / . exp. Zool. 125, 541-573.
WEISS, P. (1961). Guiding principles in cell locomotion and cell aggregation. Expl Cell Res.
Suppl. 8, 260-281.
WESTON, J. A. (1970). The migration and differentiation of neural crest cells. Adv. Morph. 8,
41-114.
WESTON, J. A. & BUTLER, S. L. (1966). Temporal factors affecting localization of neural crest
cells in the chicken embryo. Devi BioL 14, 246-266.
LOFBERG,
YANG, J., RICHARDS, J., BOWMAN, P., GUZMAN, R., ENAMI, J., MCCORMICK, K., HAMAMQTO,
S., PITELKA, D. & NANDI, S. (1979). Sustained growth and three-dimensional organization
of primary mammary tumor epithelial cells embedded in collagen gels. Proc. natn. Acad.
Sci., U.S.A. 76, 3401-3405.
(Received 28 July 1980, revised 18 November 1980)