Developmental biology

Raising the roof

Article metrics

In vertebrate embryos, the basic plan of the central nervous system forms as a precise network of neuronal cell types that are generated in ordered patterns in the developing neural tube. Two coordinated but partially independent patterning mechanisms assign positions along the anteroposterior and dorsoventral neural axes1,2. So where a particular cell is generated defines its developmental potential.

Neuronal patterning along the dorsoventral axis appears to involve specific groups of cells or organizing centres that send out opposing signals to tell neighbouring populations what kind of cells to become. Ventral signalling centres such as the floor plate and notochord control motor-neuron patterning. Now, two reports by Lee et al.3 and Millonig et al .4 on pages 734and 764 of this issue provide new insight into the specification of the most dorsal neurons. Both confirm that the roof plate, a small group of midline cells, produces signals that induce dorsal types of neurons (Fig. 1). Using elegant genetic ablation and cell-fate-mapping experiments, Lee et al.3 show that selective removal of the roof plate in mouse embryos results in loss of the most dorsal interneurons. Millonig et al.4 have identified the Lmx1a gene as the cause of defective roof-plate formation and dorso-ventral patterning in the mouse mutant dreher. These studies raise a number of questions and concentrate new interest on the roles of the roof plate in dorsal patterning and neurogenesis.

Figure 1: Dorsal signalling in the vertebrate neural tube.

a, Bone morphogenetic proteins (BMPs) derived from the epidermal ectoderm act on the neural plate to induce first neural crest and then roof plate. The roof plate in turn uses BMPs to signal to the neural ectoderm to induce dorsal types of interneurons (D1 to D3). b, Roof-plate ablation in Gdf7–DTA embryos results in absence of D1 and D2 neurons, and D3 interneurons expand dorsally3. BMP signalling from the epidermal ectoderm is not sufficient to induce these types of neurons. c, In dreher mutants there is also no roof plate, but only the specification of D1 interneurons is partially lost4. Neural crest derivatives are also affected. It is unknown whether roof-plate signals are lost completely in dreher mutants or whether there is a means of compensating for their loss in these mice. n, notochord.

The establishment of dorsoventral polarity of the neural tube depends on a ventral organizing centre. The notochord, a mesodermal rod that runs beneath the neural tube, induces overlying ventral neural tissue to form the floor plate, which can itself cause nearby cells to become ventral neurons. This action is mediated by the signalling molecule Sonic hedgehog (Shh), which is expressed first by the notochord and then by floor-plate cells. A mechanism for dorsal specification has remained more elusive, leading to the hypothesis that dorsal fates might be a default state in the absence of ventral induction. The finding that the non-neural epidermal ectoderm, adjacent to the neural plate, is responsible for the induction of neural crest and dorsal phenotypes in the neural tube5,6,7 has helped to change this view. The work of Lee et al.3 and others8,9 provides evidence for a sequential inductive pathway, whereby the epidermal ectoderm has an initial role in dorsal patterning and neural-crest formation and also induces the roof plate, which then goes on to induce dorsal neurons.

Members of the bone morphogenetic protein (BMP) family of secreted proteins are implicated in mediating signalling by both the epidermal ectoderm and the roof plate, and these tissues express different combinations of BMPs over time8. By replacing one copy of the roof-plate-restricted Gdf7 gene (itself a BMP family member9) by the gene for the diphtheria toxin A (DTA) subunit, so that the toxin would be expressed only in roof-plate cells, Lee et al.3 genetically ablated the roof plate. This caused the loss of two types of dorsal interneurons, called D1 and D2, and an expansion of more ventral D3 cell types. Lineage analysis of the Gdf7-expressing cells shows that the dorsal neurons are not descendants of the roof plate, so they are not missing simply because their progenitors have been eliminated. This has uncovered a non-autonomous role for signals from the roof plate in inducing specific groups of dorsal neurons.

The analysis of the dreher mutant adds a new player to this story, moving from outside the cell into the nucleus. Millonig et al.4 identify Lmx1a as the gene mutated or lost in three dreher alleles. There are roof-plate defects in the dreher mutant, although they are much weaker than in Gdf7–DTA embryos. The D2 interneurons are unaffected, D1 interneurons are only partially reduced and the roof-plate defects are restricted to limited regions of the anteroposterior axis. These differences in the severity of dorsal defects in dreher might result from residual roof-plate signalling in the affected regions themselves or from the spread of the signal from adjacent unaffected regions. The dreher allele studied by Millonig et al. contains only a single amino-acid replacement, so it could still have some (albeit reduced) activity. A complete null mutation or other alleles10 might have a greater effect on dorsal specification. It will be interesting to investigate how the Lmx1a and BMP signalling pathways interact.

How does specification of the roof plate itself relate to that of other dorsally derived structures, such as the neural crest and the rhombic lip (the source of the granule cells of the cerebellum)? The neural crest is normally determined before induction of the roof plate, during a narrow time window5. Lee et al.3 have shown that many neural-crest derivatives are generated in Gdf7–DTA embryos and that epidermal ectoderm maintains its ability to induce dorsal character even when the roof plate is defective. This implies that epidermal ectoderm can induce roof plate and other dorsal phenotypes (neural crest6) through BMP signalling7,8,11, but it does not normally induce differentiated neuronal phenotypes on its own — this job is reserved for the roof plate.

These studies imply that dorsal specification in the neural tube is achieved in two independent and sequential events (Fig. 1). BMP-mediated signals from the epidermal ectoderm to the neural plate induce a general dorsal character in the neural plate that would include first neural crest and then roof plate. In the second phase, the generation of different dorsal cell types would be under the specific influence of the roof plate and later the rhombic lip, independent of the epidermal ectoderm. BMPs act in both of these processes, but the question remains whether they are different mechanisms, or whether the roof plate acts as a relay centre to pass on the epidermal signal. Evidence that functions are divided between different genes, and are not part of a general combinatorial cascade, arises from the observation that only a specific subset of D1 interneurons are affected in the Gdf7 mutant mice9. Finding out which individual BMPs, receptors and associated proteins do the job in each case will be of great interest.

Finally, what other dorsal and ventral signalling centres interact with the floor plate and roof plate? The fact that not all dorsal interneurons are affected by roof-plate ablation (the D3 interneurons are still present in Gdf7–DTA mice3) indicates that there must be additional inputs into dorsal patterning. We will have to wait for a compound Gdf7–DTA (or dreher null)/Shh mutant to see what happens in the absence of both dorsal and ventral organizers, if there is anything left at all.


  1. 1

    Lumsden, A. & Krumlauf, R. Science 274, 1109–1115 (1996).

  2. 2

    Tanabe, Y. & Jessell, T. M. Science 274, 1115–1123 (1996).

  3. 3

    Lee, K. J., Dietrich, P. & Jessell, T. M. Nature 403, 734– 740 (2000).

  4. 4

    Millonig, J. H., Millen, K. J. & Hatten, M. E. Nature 403 764– 769 (2000).

  5. 5

    Selleck, M. A. J. & Bronner-Fraser, M. Development 121,525–538 ( 1995).

  6. 6

    Dickinson, M. E., Selleck, M. A. J., McMahon, A. P. & Bronner-Fraser, M. Development 121, 2099–2106 (1995).

  7. 7

    Liem, K. F. Jr, Tremml, G., Roelink, H. & Jessell, T. M. Cell 82,969–979 (1995).

  8. 8

    Liem, K. F. Jr, Tremml, G. & Jessell, T. M. Cell 91, 127– 138 (1997).

  9. 9

    Lee, K. J., Mendelsohn, M. & Jessell, T. M. Genes Dev. 12, 3394– 3407 (1998).

  10. 10

    Bergstrom, D. E., Gagnon, L. H. & Eicher, E. M. Genomics 59, 291–299 (1999).

  11. 11

    Alder, J., Lee, K. J., Jessell, T. M. & Hatten, M. E. Nature Neurosci. 2, 535–540 (1999).

Download references

Author information

Correspondence to Miguel Manzanares or Robb Krumlauf.

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.