This preview shows page 1. Sign up to view the full content.
Unformatted text preview: © 2003 Nature Publishing Group http://www.nature.com/natureneuroscience NEWS AND VIEWS
better incorporate the findings that have
recently revised our concept of radial glial
cells13–15. Only live observations—preferably
in vivo—will enable researchers to piece
together into a coherent model the data on
different modes of neurogenesis and how
these impinge on the way that neurons
migrate. With such knowledge, we may also
be better equipped to understand and eventually treat neuronal migration disorders. 1. Ross, E.M. & Walsh, C. Annu. Rev. Neurosci. 24,
2. Gleeson, J.G. et al. Cell 92, 63–72 (1998).
3. Corbo, J.C. et al. J. Neurosci. 22, 7548–7557
4. Bai, J. et al. Nat. Neurosci. 6, 1277–1283 (2003).
5. Behrends, A. & Aguzzi, A. Trends Neurosci. 25,
6. Olson, E.N., Arnold, H.-H., Rigby, P.W.J. & Wold, B.J.
Cell 85, 1–4 (1996).
7. Francis, F. et al. Neuron 23, 247–256 (1999).
8. Gleeson, J.G., Lin, P.T., Flanagan, L.A. & Walsh, C.
Neuron 23, 257–271 (1999). 9. Dykxhoorn, D.M., Novina, C.D. & Sharpe, P.A. Nat.
Rev. Mol. Cell Biol. 4, 457–467 (2003).
10. Frantz, S. N at. Rev. Drug Dis. 2 , 763–764
11. Marin, O. & Rubenstein, J.L. Annu. Rev. Neurosci.
26, 441–483 (2003).
12. Miyata, T., Kawaguchi, A., Okano, H. & Ogawa, M.
Neuron 31, 727–741 (2001).
13. Morest, D.K. & Silver, J. Glia 43, 6–18 (2003).
14. Malatesta, P. et al. Neuron 37, 751–764 (2003).
15. Gupta, A. et al. Nat. Neurosci. 6, 1284–1291
(2003). Specifying motor neurons: up and down and back to
William A Harris
A recent study in Nature shows that the columnar fate of motor neurons in the embryonic spinal cord is imposed by crossrepressive patterns of Hox-c expression. This Hox expression is in turn controlled by graded FGF signaling. During development, motor neurons are
organized along the head-to-tail or rostralcaudal axis of the spinal cord into distinct
columns, which innervate different targets1.
The cervical cord, for example, has a lateral
motor column (LMC) that innervates the
forelimb, whereas the thoracic cord has a
distinct set of autonomic motor neurons in
the Column of Terni (CT). Because the
spinal motor neurons come from progenitor cells in the same ventral region of the
neural tube, the columnar organization is
thought to come from an independent rostral-caudal patterning mechanism. In a
stunning study in Nature, Dasen, Liu and
Jessell2 now show that the rostral-caudal
patterning of motor neurons into columns
occurs in response to a gradient of fibroblast growth factor (FGF). This gradient
establishes domains of a family of homeobox (Hox) genes, which are expressed at
different levels; the boundaries between
these domains then establish the boundaries of the different motor pools.
To appreciate how FGF signaling could be
working to establish motor columns, it helps
to think about how the spinal cord develops
in the embryo. During gastrulation, cells
involute through the primitive streak, a strip
of cells that defines the rostral-caudal axis of
the embryo. The site of involution is called
the node, and as gastrulation proceeds, the William Harris is at the Department of Anatomy,
Cambridge University, Downing Street, Cambridge
CB2 3DY, UK.
e-mail: firstname.lastname@example.org node regresses caudally along the streak, leaving more mature involuted mesodermal cells
more rostrally. Near the node is the presomitic mesoderm, which expresses FGF8,
and which now directly underlies the forming neural plate3. Two recent papers in
Neuron show that FGF inhibits neuronal differentiation at this stage, and so the cells in
the overlying neural plate remain as dividing
Previous transplant studies in chicks show
that the transposition of mesoderm from
thoracic to cervical levels changes the fate of
the overlying motor neuron columns from
LMC to CT identities5. Moreover, previous
studies6,7 show that caudal mesoderm
expresses higher levels of FGF8 than does
rostral mesoderm, and that the low levels of
FGF induce the expression of rostral Hox
genes such as Hoxc5 and Hoxc6, whereas high
levels of FGF induce the expression of the
more caudal genes Hoxc8 and Hoxc9. Thus,
FGF8 released from the presomitic mesoderm, although it keeps motor neurons from
differentiating, also seems to be involved in
patterning the rostral-caudal axis of the
spinal cord. The simple hypothesis is that
FGF8 regulates the pattern of Hox genes,
which in turn specify motor column identity.
Dasen et al.2 now provide direct evidence
in support of this hypothesis, by reporting
that the progenitor cells that give rise to
motor neurons express Hox proteins in
response to graded FGF signaling, and that
motor neuron columnar fate develops
through the transcriptional functions of the
Hox proteins. In embryonic chick spinal
cord, they found that Hox protein expression N ATURE NEUROSCIENCE VOLUME 6 | NUMBER 12 | DECEMBER 2003 segregated with motor neuron columnar
subtype. For instance, Hoxc6 expression was
confined to cervical levels of the embryonic
spinal cord, whereas Hoxc9 expression was
restricted to thoracic levels. When FGF8 signaling was increased in the neural tube, this
caused a repression of Hoxc6 expression and
the onset of Hoxc9 expression at cervical levels. Moreover, if Hoxc9 was misexpressed cervically, it repressed the expression of Hoxc6,
and when Hoxc6 was misexpressed thoracically, it repressed the expression of Hoxc9,
showing that there is mutual inhibition,
leading to a sharp border between Hoxc6and Hoxc9-expressing motor neurons.
No neurons expressed both Hoxc6 and
Hoxc9. Similar mutual repression was found
between Hoxc5 and Hoxc8. Hoxc6-expressing motor neurons, whether they were in the
normal cervical position or the abnormal
thoracic position, showed characteristics of
the LMC, such as expression of the retinoic
acid synthesizing enzyme Raldh2, axonal
projections toward the limb bud, and an
ovoid cross-section in the cord. Similarly,
Hoxc9-expressing neurons, whether they
were in their normal thoracic position or in
the cervical region (due to treatment with
extra FGF or misexpression), showed CT
characteristics such as BMP5 expression,
axonal projection to the sympathetic chain
and a medial semi-lunar cross-section in the
cord. Thus an FGF gradient establishes a
motor neuron Hox pattern that determines
Developmental neurobiologists will appreciate that this logic is strikingly similar to that
governing the positioning of the motor neu- 1247 © 2003 Nature Publishing Group http://www.nature.com/natureneuroscience NEWS AND VIEWS
Figure 1 Motor neuron specification in the
dorsoventral and rostrocaudal axis of the spinal
cord. The spinal cord diagram (b) shows the
ventral location of the motor neurons and two
different motor columns that arise at different
rostrocaudal levels of the cord. Ventral to the
cord is the medial notochord, and the paraxial
mesoderm. (a) A gradient of Shh emanates from
the notochord and floor plate; threshold levels of
Shh turn on Class-II HD genes. Retinoic acid (RA)
expressed by the paraxial mesoderm induces the
expression of Class-I HD genes that are turned off
more ventrally by threshold levels of Shh. Class-I
and Class-II HD transcription factors cross
repress each other, creating sharp definitive
boundaries at different dorsoventral levels in the
cord. Thus the boundary between Dbx and Nkx6
is more dorsal than the boundary between Pax6
and Nkx2.2. Between these boundaries, the
OLIG2 bHLH transcription factor necessary for
motor neuron specification is turned on by the
concerted action of RA, Nkx6 and Pax6. OLIG2
and RA are necessary for the expression of motor
neuron identity genes in motor neuron precursors
(pMN). (c) A gradient of FGF8 emanates from the mesoderm. High levels of FGF8 turn on more caudal Hoxc genes, whereas RA and low levels of FGF8 turn
on rostral Hoxc genes. Rostral and caudal Hoxc transcription factors cross-repress each other, creating sharp definitive boundaries at different rostrocaudal
levels in the cord. The boundary between Hoxc6 and Hoxc9 establishes the boundary between the LMC of the cervical cord and the CT of the thoracic cord.
The significance of the Hoxc5 and Hoxc8 boundary is unknown. rons in the ventral region of the spinal cord,
which the Jessell laboratory has demonstrated
in a series of exceptional papers (reviewed in
ref. 8). Dorsal-ventral patterning in the spinal
cord begins in response to a gradient of Sonic
hedgehog (Shh) protein secreted by the notochord and then the floor plate. This gradient
sets up expression domains of two kinds of
homeodomain (HD) transcription factors:
Class-I HD factors (like Nkx2.2 and Nkx),
which are turned on in the ventral cord by high
levels of Shh, and the more dorsal Class-II HD
factors (Pax6 and Dbx2), which are turned
off by high levels of Shh in the ventral cord.
Class-I and Class-II HD genes are differentially
sensitive to Shh and form cross-inhibitory
relationships such that sharp borders of
expression are set up at different dorsoventral
levels of the spinal cord. These borders define
discrete dorsal-ventral domains of the cord,
including the domain of motor neurons. The
domain that expresses both the Class-I HD
gene Nkx6 and the Class-II HD gene Pax6 also
turns on OLIG2, a bHLH transcription factor
that is required for motor neuron differentiation. OLIG2 cells then turn on motor neuron
identity genes such as Mnx and Lim3, initiating
the program of motor neuron differentiation
in this domain (Fig. 1).
Novitch et al.3 and Diez et al.4 show that in
addition to Shh, FGF8 and retinoic acid (RA)
are also involved in the dorsal-ventral patterning of the spinal cord3,4, and RA contributes to
establishing the rostral-caudal pattern of
motor columns9. Thus, FGF8 and RA are 1248 involved in both the rostral-caudal and dorsalventral specification of motor neurons. To
understand how this works, let us return to the
developing embryo. At mid-gastrulation, the
node has proceeded half way down the primitive streak, creating a rostral-to-caudal gradient of differentiation. Rostrally, the mesoderm
that involuted through the node earlier has
organized itself into a ventromedial notocord
and a lateral somitic or paraxial mesoderm. At
this stage, FGF8 is turned off in the mesoderm, which allows the differentiation of
motor neurons in the overlying neural plate,
now curling itself into a neural tube that will
form the spinal cord3,4. At this time also, the
paraxial mesoderm begins secreting RA, while
the notochord begins secreting Shh3. RA
inhibits FGF8 expression in the somitic mesoderm. Similarly FGF turns off RA expression4.
This reinforces a pattern in which posterior
somatic mesoderm is higher in FGF signaling
while more anterior presomitic mesoderm is
higher in RA. In the anterior region, high RA
collaborates with low FGF in turning on the
neural differentiation program4. Sockonathan
et al.9 now show that RA also collaborates with
FGF in columnar specification. Cervical
motor neurons expressing dominant-negative
RA receptors acquire the phenotypes of CT
motor neurons, whereas expression of constitutively active RA receptors in thoracic regions
prevents the normal differentiation of CT
How do these signaling systems contribute
to the transcriptional cascade involved in the dorsal-ventral positioning of motor neurons?
By exploring how FGF inhibits motor neuron
differentiation, it was discovered that FGF
signaling inhibits the expression of both ventral Class-I and dorsal Class-II HD transcription factors in the neural tube3,4. Shh, as we
know, is expressed by the notochord as FGF
signaling is turned off, and turns on the ventral Class-II HD genes8. But how do the
Class-II HD genes get turned on? RA is
expressed more dorsally by the paraxial
mesoderm and can activate Class-II HD
genes3. Thus, the combination of RA and Shh
signaling turns on both the Class-I and ClassII HD gene. Interestingly, most of the transcription factors in the dorsal-ventral cascade
work as repressors. For example, Nkx6 and
Pax6 do not directly induce OLIG2, OLIG2 as
a repressor cannot directly induce motor
neuron identity genes. In both of these steps,
RA signaling is required, as dominant-negative RA receptors misexpressed at either stage
block progression of the cells toward the
motor neuron fate. RA might therefore act as
a de-repressive signaling system needed to
inhibit the repressors in this essentially
repressive transcriptional network3.
This new work leaves us with a much
richer view of how the three-dimensional
system of motor neurons develops in an animal by the coordinated time-dependent,
concentration-dependent and positiondependent actions of FGF, RA and Shh. That
FGF8 and RA are involved in both dorsalventral and rostral-caudal patterning may VOLUME 6 | NUMBER 12 | DECEMBER 2003 N ATURE NEUROSCIENCE © 2003 Nature Publishing Group http://www.nature.com/natureneuroscience NEWS AND VIEWS
only be possible because they are working
during embryogenesis, a kind of timewarped space where the relative position of
signaling centers and responding cells
change as a function of developmental time.
These findings strongly resemble limb morphogenesis, in which a complex timedependent integration of BMP, RA, Shh and
FGF signaling within rostral-caudal, dorsalventral and proximal-distal axes create distinct Hox protein patterns that specify bone
and digit identity.
The new data also raise intriguing new
questions such as how exactly RA-mediated
de-repression works. The dissociation of CT and LMC columns is only the beginning
of a dissection of several different motor
columns and subdivisions of motor columns
that have overlapping yet distinct locations
along the extent of the spinal cord. Still, the
recent work has added a new dimension to
the work on motor neuron development, by
showing that the principles of homeodomain protein function along the dorsalventral axis and rostral-caudal axis are
similar—a graded activity of a signaling factor establishes homeodomain boundaries,
which are further refined by selective crossrepression among the homeodomain transcription factors. The convergence of these two programs then ensures correct motor
neuron specification along the dorsal-ventral and rostral-caudal axes.
1. Tosney, K.W., Hotary, K.B. & Lance-Jones, C.
Bioessays 17, 379–382 (1995).
2. Dasen, J., Liu, J.-P. & Jessell, T. Nature 425,
3. Novitch, B.G., Wichterle, H., Jessell, T.M. &
Sockanathan, S. Neuron 40, 81–95 (2003).
4. Diez del Corral, R. et al. Neuron 40, 65–79 (2003).
5. Ensini, M., Tsuchida, T.N., Belting, H.G. &
Jessell, T.M. Development 125, 969–982 (1998).
6. Liu, J.P., Laufer, E. & Jessell, T.M. Neuron 32,
7. Bel-Vialar, S., Itasaki, N. & Krumlauf, R.
Development 129, 5103–5115 (2002).
8. Jessell, T. M. Nat. Rev. Genet. 1, 20–29 (2000).
9. Sockanathan, S., Perlmann, T. & Jessell, T.M. Neuron
40, 97–111 (2003). Folding the cortex
The cortical surface of a normal mouse is smooth
(top). The cortices of primates, including humans, are
folded up into a complex landscape of gyri and sulci.
In this issue (page 1292), Kingsbury et al. report that
treating cultured cortical hemispheres from mouse
embryos with lysophosphatidic acid (LPA) for only 17
hours caused them to expand and fold up into a structure that resembles a primate brain (bottom; blue is
The folded cortices were thicker and contained
more cells than controls: both more differentiated
neurons in the cortical plate and more neural progenitor cells in the ventricular zone. LPA induced terminal
mitosis—the generation of neurons from progenitors—and it also reduced apoptotic death of progenitor cells. Although LPA can induce cell proliferation in
other systems, Kingsbury et al. saw no increase in
S-phase cells, so most of the increase in cell number
seemed to be due to inhibition of cell death. The
effects of LPA were mediated by one or both of the
closely related G protein–coupled receptors LPA1 and
LPA2, as cortices from mouse embryos lacking these
two receptors did not fold up after LPA treatment.
Neurons in the brain produce LPA, which can activate a number of intracellular signaling pathways.
Despite the presence of the LPA signal, however, mice
lacking LPA1, LPA2 or both receptors show no evident
cortex malformation, suggesting that such signals are
not important for the development of the smooth cortex of the mouse. On the other hand, the ability of
excess LPA to induce cortical folding in mice suggests
the intriguing possibility that LPA might contribute to
cortical development in primates. Annette Markus NATURE NEUROSCIENCE VOLUME 6 | NUMBER 12 | DECEMBER 2003 1249 ...
View Full Document
- Fall '10