. 27
( 66 .)


can link up with each other using long, tenuous, ¬lopodial exten-
sions known as cytonemes (Ramirez-Weber and Kornberg, 1999; Bryant,
1999), which can form gap junctions at their sites of contact or thin
nanometer-diameter membrane tethers (Rustom et al., 2004). Mes-
enchymal cells can also communicate across the ECM using diffusible
molecules (˜˜paracrine” signaling), a topic discussed at length in Chap-
ter 7. Most interestingly from the physical viewpoint is the possibility
that mesenchymal cells communicate over long distances mechani-
cally. That is to say, cells could exert a force on their local environ-
ment and, if there is a suf¬cient density of interconnections within
this environment, other cells could experience this force. We will take
up this subject and, in particular, the rigorous physical meaning of
˜˜suf¬cient density of interconnections,” in this chapter.
In earlier chapters the physical and mathematical models we dis-
cussed were applied directly to developmental processes taking place
in living systems: embryos or their tissues. In our overview of the
viscoelastic and network (i.e., interconnection) properties of ECMs we
will ¬nd it useful to present results from some simpli¬ed nonliv-
ing experimental models. One of these is puri¬ed type I collagen, the
most abundant protein of the ECM and, for that matter, of the animal
kingdom, undergoing assembly (Newman et al., 1997; Forgacs et al.,
2003). We also describe an in vitro phenomenon known as ˜˜matrix-
driven translocation” (MDT), which is a morphogenetic rearrangement
seen in suspensions of cells or certain types of latex particles dis-
persed in puri¬ed ECM components (Newman et al., 1985; Forgacs
et al., 1989). The usefulness of such an approach lies in the possi-
bility of accounting for complex behaviors of living tissues on the
basis of the interactions of a reduced number of their components.
Thus the spirit of this approach is similar to that of the idealized
mathematical--physical models discussed earlier.
The phenomenon of mesenchymal condensation is important in many
events of early embryogenesis and organ formation (Hall and Miyake,
1995, 2000; Newman and Tomasek, 1996). This is usually a transient ef-
fect in development in which mesenchymal cells, initially dispersed

in a matrix, move closer to one another. Condensations generally
progress to other structures, such as feather germs (Chuong, 1993),
cartilage or bone (Hall and Miyake, 1995; 2000), and kidney tubules
(Ekblom, 1992). It is clear why condensation is characteristic of mes-
enchymal, but not epithelioid, tissues; the cells in the latter are al-
ready as close together as they can be. It is helpful to recognize that
since ˜˜epithelial” and ˜˜mesenchymal” refer to the physical states of
tissues, the conversions mentioned can go in either direction dur-
ing development: epithelial to mesenchymal during gastrulation, or
mesenchymal back to epithelioid tissue, transiently, as in precartilage
condensations, or permanently, as in kidney tubules.
In the following we will ¬rst describe the morphogenesis of mes-
enchyme in the vertebrate neural crest. This will be followed by a
discussion of the MDT experiment and its possible relevance to this
and other mesenchymal rearrangements in the embryo. Because the
MDT phenomenon suggests that mesenchymes act as coherent ¬‚uids,
as a result of the network properties of their ECMs, we next present
an analysis (based mainly on the concept of percolation) of network
formation during collagen assembly. Subsequently we will describe
mesenchymal condensation in several embryonic systems, with em-
phasis on the precartilage condensations of the developing vertebrate
limb. Finally, we will bring together a number of the physical concepts
from this and preceding chapters and introduce a physical model for
the formation of such condensations.

Development of the neural crest
The neural crest consists of populations of cells that detach and mi-
grate away from the dorsal ridges (referring to the embryo™s back sur-
face) of the neural tube just before, or shortly after it closes (Langille
and Hall, 1993). The detachment of neural crest cells from the neural
tube must involve the loss or modi¬cation of cell--cell adhesion, but
which molecules and mechanisms may be involved is not yet clear.
At least two Ca2+ -dependent cell--cell adhesion molecules, N-cadherin
and E-cadherin, are lost from the neural crest cells prior to, or shortly
after, detachment from the neural tube (Akitaya and Bronner-Fraser,
1992; Bronner-Fraser et al., 1992).
Other suggested mechanisms for the detachment of neural crest
cells from the neural tube are the degradation of adhesion molecules
by secreted enzymes called proteases and the generation of tractional
forces that would mechanically rip cells away from their neighbors
(Erickson and Perris, 1993). Although neural crest cells are known to
produce proteases that are associated with their increased motility in
culture (Valinsky and Le Douarin, 1985; Erickson and Isseroff, 1989),
evidence suggests that some form of epithelial-to-mesenchymal trans-
formation, based on the acquisition of new cell-surface properties
(see, e.g., Ozdamar et al., 2005), is the likeliest mode of detachment
of these cells (Newgreen and Minichiello, 1995; Duband et al., 1995).

The detachment of cells per se (which we have encountered already
in connection with the formation of primary mesenchyme in the
sea urchin embryo, see Chapter 5) is not a mechanism that calls for
novel physical explanations. But neural crest cells also translocate
enormous distances (on the cell™s scale), and the mechanisms of such
movement, like any rearrangement of matter, do suggest that some
interesting physics is at work. The cells of the neural crest give rise
to a variety of widely dispersed cell types (neural crest derivatives)
of the adult body. These include the nerve cells of the peripheral
nervous system (i.e., other than those of the brain and spinal cord),
the endocrine cells of the adrenal medulla and those of the thyroid
gland that produce the hormone calcitonin, the pigment cells of the
skin, cartilage and bone of the face, and the connective tissues of
the teeth, several different glands, and a number of arteries. A major
question for the physical understanding of development is how the
cells of the neural crest convey themselves, or are conveyed, to the
distant sites at which they will differentiate into these derivatives.
The unique nature of mesenchymal cells as a population free from
broad cell--cell attachment permits them to rearrange and translocate
by mechanisms distinct from those we have discussed for cells in epi-
thelioid masses and epithelial sheets. For example, cells could migrate
individually, moving toward sources of chemoattractants. This is re-
ferred to as chemotaxis (see Chapter 1). Another mode of translocation
of individual cells, known as haptotaxis, is related to the differential
adhesion mechanism we discussed for epithelia. Haptotaxis requires
an external gradient (i.e., a nonuniform distribution) of a substratum
adhesive material, typically a complex of ECM molecules. A randomly
locomoting cell that binds in a reversible fashion to such a molecu-
lar substratum will eventually wind up at the site that minimizes its
energy of adhesion, either locally or, if the cell™s inherent motility is
suf¬ciently strong to ˜˜kick” it past a local minimum, globally. A spe-
ci¬c physical model of haptotaxis that characterizes the conditions
under which cells may redistribute in relation to adhesive gradients
was constructed by Dickinson and Tranquillo (1993).
Although chemotaxis and haptotaxis are plausible mechanisms
for the translocation of mesenchymal cells during embryonic devel-
opment, and haptotaxis indeed appears to be employed during the
formation of mesenchymal condensations (see below), there does not
appear to be a signi¬cant role for these mechanisms in the migration
of the neural crest. One main pathway of neural crest dispersal in the
˜˜trunk” (main-body) region of the embryo takes them from their ori-
gin at the dorsal neural tube into lateral extracellular spaces between
the ectoderm and the somites (Fig. 6.2, Path 1). The somites are blocks
of mesodermal tissue that form parallel to, and to either side of, the
notochord (see Chapter 7). When labeled test cells were grafted onto
sites lateral to the neural tube, they migrated medially (i.e. toward the
neural tube) between the ectoderm and somites and ventrally (i.e.
toward the belly of the embryo -- the opposite of dorsally) along
blood vessels between the somites. These exogenous (i.e., transplanted)



Neural tube




Fig. 6.2 Pathways of dispersal of trunk neural crest cells in a vertebrate embryo. The
stage of development shown corresponds to the bottom panel of Fig. 5.13. In Path 1,
cells move through the ECM located between the ectoderm and the somites, whereas in
Path 2 they move through the anterior region of the sclerotome (the cartilage-forming
region of a somite) and enter the ECM between the sclerotome and dermamyotome
(the connective tissue and muscle-forming region of the somite). Path-1 neural crest cells
differentiate into pigment cells of the skin, while path-2 cells differentiate into neurons,
the medullary (interior) cells of the adrenal gland, and Schwann cells, the insulating cells
of the nervous system. The aorta is the main artery leading away from the heart (see
Chapter 8).

neural crest cells thus moved in a direction opposite to that of the
endogenous (i.e., normally present) trunk neural crest cells. The other
trunk neural crest pathway extends ventrally along the space that sep-
arates the portion of each somite that will form bone from the portion
that will form connective tissue and muscle (the ˜˜intrasomitic space”)
(Fig. 6.2, Path 2). When neural crest cells were grafted onto this path-
way they migrated rapidly, within 2 hr, in two directions: dorsally,
eventually contacting the ventrally moving stream of host neural crest
cells, as well as laterally. These experiments indicate that neither a
preestablished chemotactic nor an adhesive (haptotactic) gradient ex-
ists in the embryo, since the grafted neural crest cells will move in
the reverse direction from normal along these pathways, toward the
dorsal neural tube (Erickson, 1985; see also Erickson, 1988).
Such puzzling results have focused attention on the ECM through
which the neural crest cells move, which, along with the cells, con-
stitutes a material that could potentially exhibit ¬‚uid properties. For
example, the connective tissues of the vertebrate head are largely
the product of the cephalic (also termed cranial) neural crest. These
cells arise from the dorsal-most regions of the future brain and mi-
grate laterally along several pathways (Noden, 1984). Cephalic neural
crest migration in the embryo of the axolotl (an amphibian) follows
a ˜˜micro¬nger” morphology consisting of several parallel streams
(H¨rstadius and Sellman, 1946) (Fig. 6.3A). Previously Noden had
suggested that the invading sheet of neural crest cells might be

Fig. 6.3 (A) Patterns of dispersion of normal and heterotopically grafted (i.e., to an
abnormal site) neural crest cells in the axolotl embryo. On the left, the right neural ridge
of the head has been stained with neutral red. The left ridge has been excised, stained
with Nile blue, and implanted lower down on the same side. On the right, streams of
neural crest cells from the right neural ridge (red) move downwards (the normal
migration pathway for this cell population) to where they meet streams of cells (blue)
migrating from the graft in the dorsal direction. The whole early embryo and the portion
shown of the later embryo are each about 2 mm in width. (B) The time evolution of
structured ¬‚ows in a polymer solution containing dextran and polyvinyl pyrrolidone
(PVP). PVP in the lower layer was coupled to a dark blue dye that appears black in the
¬gure. On the left, the initial layered-¬‚uid preparation. On the right, the ¬‚uid
con¬guration after 40 min. Micro¬ngers of width on the order of 500 µm have formed
by the oppositely directed vertical ¬‚ow of the polymer solution. Note the striking
similarity between the micro¬ngering patterns in A and B. (A, based on Horstadius and
Sellman, 1946; B, courtesy of Dr Wayne D. Comper; see Comper et al., 1987. The entire
¬gure is based on Newman and Comper, 1990.)

divided into streams by anatomical obstructions encountered in their
path (Noden, 1988). However, neural crest cells transplanted into ven-
tral regions of axolotl embryos break into micro¬ngers of similar
dimension as they migrate dorsally through a region of the embryo,
distant from the endogenous streams (H¨rstadius and Sellman, 1946),
that is unlikely to contain an equivalent set of obstacles. This led Com-
per et al. (1987) to suggest that such micro¬ngering ¬‚ows through
the ECM might be based on a physical transport mechanism that had
been observed in nonuniform solutions of polymers (Preston et al.,
1980). Such convective ¬‚ows, which take the form of interdigitating
micro¬ngers (Fig. 6.3B), were produced in solutions of ECM proteo-
glycans (see Chapter 4) (Harper et al., 1984) or collagen (Ghosh and
Comper, 1988).

In this example, a physical system, consisting of polymers de-
rived from tissues, exhibited a morphological outcome reminiscent
of streams of neural crest cells. More direct evidence that physical,
as opposed to cell-motility-driven, ¬‚ow processes are involved in neu-
ral crest dispersion comes from experiments performed by Bronner-
Fraser and coworkers (Bronner-Fraser, 1982, 1984, 1985; Coulombe
and Bronner-Fraser, 1984). In these studies, cell-sized polystyrene latex
beads were microinjected into the ventral neural crest pathway of the
trunk region of avian embryos. It was found that the beads were able
to translocate along this pathway, accumulating in the vicinity of en-
dogenous neural crest derivatives (Bronner-Fraser, 1982). This occurred
even if the host neural crest cells were ablated by a laser, indicating
that the beads were not being passively conveyed by migrating cells
(Coulombe and Bronner-Fraser, 1984).
Although nonmigratory, nonneural crest cells also translocated
along the ventral trunk pathway when microinjected (Bronner-Fraser,
1982), latex beads coated with the ECM proteins ¬bronectin or
laminin (both negatively charged) were restricted from entering the
pathway. Charge, however, was not the determining factor: uncoated
beads, and beads coated with type I collagen, are also negatively
charged and they both translocated ventrally after injection (Bronner-
Fraser, 1984). And while beads coated with polylysine, which renders
them positively charged, were not capable of being translocated along
the ventral neural crest pathway, polytyrosine-coated beads, with no
surface charge at all, were translocated (Bronner-Fraser, 1984).
These intriguing experiments strongly suggest that the movement
of actual cells along neural crest pathways does not depend on their
intrinsic motility or on any obvious speci¬c feature of their surfaces,
just as it does not depend on chemotactic or haptotactic gradients
(as we have seen above). Such evidence inevitably raises the possibility


. 27
( 66 .)