The observation that incited us to launch the present study is illustrated in figure 1. At around E17.5, and compared to other regions of the brain, a relatively high level of the S100B-EGFP reporter protein was found in the Cb primordium (Fig 1A), and this was matched by a fair level of S100B protein expression in densely packed radial glial cells located at the midbrain-hindbrain boundary (Fig 1B, MHB) or bordering the fourth ventricle (Fig 1C), in isolated cells of the cortical transitory zone (Fig 1D), and to a lesser extent in the RL (Fig 1E). Unlike EGFP, both nuclear and cytoplasmic, the S100B signal was often restricted to the cytoplasm but specificity was inferred from the total absence of S100B immunosignal in corresponding sections obtained from S100B null embryos (data not shown).

Between E13.5 and E17.5, migratory Purkinje progenitors are subject to "contact guidance", a process by which they adhere to radial glial processes to reach their final destination in the cortex 26. From E14.5 until P7, Bergmann glial cells are thought to derive from primary radial glial cells that translocate their somata from the VZ into the Purkinje cell layer, giving rise to a secondary radial glial scaffold 27. Detection of S100B in the cerebellar radial glial scaffold at E17.5 therefore incited us to go back in time and determine exactly when, and in which cell type, transcription of the S100B gene is activated in the embryonic Cb.

Compared to E17.5, the number of EGFP+ VZ cells co-expressing S100B was much reduced at E13.5, except in the narrow strip of cells located at the midbrain-hindbrain boundary (compare the boxed areas in Figs. 1A and 2A,B). To better understand the 3-dimentional structure of this region, we examined coronal sections as well and found that EGFP expression was prominent in a medial stretch of ventral neuroepithelium of approximate width 300 microns that belongs to the inferior tectal neuroepithelium, and not to the Cb as originally thought (Fig 2c). In the Cb, the number of EGFP+ cells was strikingly increased near the midline and organized in an S-shaped intensely EGFP+ radial glial scaffold spanning the primordium from the VZ to the pial surface (Fig. 2E). The scaffold could only be visualized in its integrity in paramedian sections (compare Figs 2D and 2F).

Figure 3A illustrates the general pattern of EGFP expression observable on parasagittal sections of a Cb hemisphere at E14.5. The radial glial nature of the cells aligned in the VZ was first inferred from the oval-shaped somata and long ascending processes traversing the entire thickness of the Cb primordium filled with post mitotic β 3 tubulin+ neurons (Fig 3B, inset). The radial glial nature of the cells was confirmed using the technique of immunohistochemistry to detect the expression of two markers: brain lipid binding protein (BLBP), a member of the large family of hydrophobic ligand binding proteins exclusively expressed in radial glia during embryonic brain development 28, and Sox9, a transcription factor expressed in the neuroepithelium and choroid plexus epithelium 29. The high level of EGFP;S100B and EGFP;BLBP co-expression is illustrated in figures 3C, and 3D, respectively. As expected, we found that epithelial cells of the choroid plexus, radial glial cells and isolated cells located in the CTZ also expressed the transcription factor Sox9 from E13.5 (not shown) to E16.5 (Fig 3E).

Together these results indicated that before midline fusion of the Cb plates, EGFP and S100B were co expressed in the Sox9+ BLBP+ primary radial glial scaffold traversing the Cb primordium from the IC posterior limit to the RL, with a higher level of expression near the midline. In addition, EGFP;S100B co-expression was prominent in a median stretch of inferior tectal neuroepithelium directly abutting the rostral Cb primordium.

In the mouse, all cerebellar PCs are produced during only three embryonic days from E10.5 to E12.5 30, and most, if not all glial precursors in the Cb take the form of radial glial cells from E12.5 to E14.5, during the phase of PC migration 27. Using an antibody directed against calbindin-1 (spot 35), which starts being expressed in post-mitotic PCs 26, we found no colocalisation of calbindin and EGFP from E13.5 until E16.5 (Fig. 4). Beginning on E13.5, EGFP expression was restricted to the single cell layer of radial glia bordering V4, whereas calbindin+ PCs were distributed in a broad cellular cortical zone of the future hemisphere (Fig. 4A). Some PCs located near the VZ were closely apposed to EGFP+ radial processes (Fig. 4B), which is a characteristic feature of migrating progenitors. The number of EGFP+ cells was greatly increased at E16.5 but EGFP and calbindin were never co-expressed in the same cells (Fig. 4C–F).

Based on the evenly distribution and morphology of EGFP+ cells connected to the pial surface via their apical radial process (Fig. 4E,F), which characterize prospective Bergmann glial cells 31, we conclude that the radial glia-derived EGFP+ population present in the Cb primordium between E13.4 and E17.5 likely contains glial precursors some of which are already contributing to the secondary Bergmann radial glial scaffold that will later be used by GPs to migrate from the EGL to the IGL.

Before E12.5, EGFP expression was not detected in the RL (Fig 5), or any other territory of the CNS. Beginning on E12.5, EGFP expression was detected in a subpopulation of bipolar cells present in the RL (Fig 6A,B) and emitting subpial branching processes into the nascent EGL (Fig 6C). However, in contrast to cells present in the medial portion of the cerebellar VZ, BLBP expression was much reduced and expression of the transgene was not matched by detection of the S100B protein (data not shown). With some exceptions 32, it is now generally accepted that the RL generates exclusively neuronal precursors. Precursors of DCN glutamatergic projection neurons are born in the prospective RL and migrate into the future EGL before gathering in the NTZ between E10 and E13.5 89, whereas the first cerebellar GPs and UBCs are born during the last gestational week beginning on E13 7. Therefore, we assumed that EGFP expression in the RL beginning on E12.5 could reflect activation of the S100B gene in GP or (and) UBC precursors, but not in DCN precursors.

The non DCN nature of EGFP+ RL precursors was tested by performing immunohistochemistry with a pair of antibodies directed against the transcription factor Pax6 expressed in the RL throughout its neurogenetic interval 10, and the transcription factor Olig2 known to label the NTZ starting on E12.5 33. The high proportion of EGFP+ Pax6+ double positives present in the lateral portion of the RL at E12.5 is illustrated in figures 6B and 6C (colored dots). Their non-DCN nature was inferred from the discontinuity that existed at E13.5, between the NTZ already filled with Olig2+ nuclei and the territory of EGFP expression restricted to the RL and posterior EGL (Fig 6D).

Until E14.5 (Fig 7), a high level of EGFP expression was maintained in a subpopulation of Pax6+ precursors present in the RL and in the EGL, whereas from E15.5 (data not shown) onwards (Fig 1A, Fig 4C–F), EGFP expression was down-regulated in the EGL, and the number of EGFP+ cells present in the RL decreased considerably in proportion to the growth of the EGL. During that time, expression of the transgene was not matched by a significant S100B signal (data not shown). However, the possibility that the transgene could be ectopically expressed was discredited once we realized that S100B mRNA had previously been detected in that region 34 (high resolution pictures of the E14.5 stage available 35).

From our results and the results provided by a recent study that will be discussed below 7, we believe that transient expression of the S100B-EGFP transgene is a specific marker of the primary phase of GP genesis in the embryonic RL.

Although we found no mention of the S100B protein being expressed in the mouse embryonic cerebellar radial glia, the S100B mRNA was detected in the rat cerebellar radial glia 12. Additionally, in a study designed to analyze the expression pattern of 158 murine orthologs of genes identified on human chromosome 21, the S100B mRNA was detected in the VZ of the mouse embryonic Cb on whole mounts and tissue sections 34.

The radial glial origin of EGFP+ S100B+ cells present in the CTZ in increasing number from E14 to E17.5 is also consistent with several reports showing that in addition to being essential for PC exit from the VZ between E13.5 and E17.5 36, the primary radial glia is linked to gliogenesis from E14 to P7 27. According to ³H thymidine and BrdU birth dating studies, PCs are produced between E11 and E13 37, and expression of calbindin-1 is first detectable at around E14 in post mitotic PCs that have migrated a short distance away from the VZ 26.

Here we show that EGFP+ cells that detach from the VZ and migrate into the CTZ beginning on E14.5, are not post-mitotic Purkinje cells. While migrating towards the surface most EGFP+ cells lose their connection to the VZ but not their connection to the pial membrane, a feature of prospective Bergmann glial cells. Finally, we show that near the onset of gliogenesis (E16.5), Sox9 expression is maintained in some radial glia-derived EGFP+ cells. This is in line with a previous report showing that most, if not all glial precursors in the Cb take the form of radial glial cells 27 and Bergmann glial cells maintain Sox9 expression from the early migrating progenitor stage through adulthood 29.

Several molecules known to be expressed in Bergmann glial cells of the adult mouse Cb start being expressed in the embryonic radial glia, at around E14. They include the intermediate filament protein vimentin 4, the lipid binding protein BLBP 38, the enzyme Glutamine synthetase 39, the extra cellular matrix molecule with neuron-glia cell adhesion activity Tenascin-C 40, the glutamate transporter Glast 41, the enzyme 3PGDH involved in L-serine biosynthesis 42, the Punc neural cell-adhesion molecule 43, and the transcription factor Sox9 29. We now suggest that S100B be added to the list.

The strongly S100B+ median stretch of ventral neuroepithelium we observed at the midbrain-hindbrain junction (Fig 3E) is reminiscent of a previously reported vimentin+ radial glial structure linking the IC to the Cb in E20 rat embryos 44. In the mouse, this isthmic territory is thought to provide a cellular substratum and the signals essential for fusion of the cerebellar plates along the dorsal midline 44. In addition, this territory is subsequently relinquished to the velum medullaris, a neuron-free sheet of cells that is very short before birth but which develops during the following days and links the anterior pole of the vermis to the inferior colliculus.

The RL is a specialized germinative epithelium that arises as a result of the ongoing interaction between the neural tube and the non-neural roof plate ectoderm of V4 (reviewed in 1). Its main derivatives include primary precursors of glutamate releasing excitatory granule neurons, by far the most numerous neurons in the CNS, which relay afferent, excitatory information from mossy fibers to Purkinje neurons. In addition, the RL produces precursors of DCN neurons, the main output centers of the Cb 9, and cortical UBCs, which amplify inputs from vestibular ganglia and nuclei 45.

In the present study, we provide evidence for transcription of the S100B gene being specifically activated in the RL during the primary phase (E13–E17) of GP genesis. First, we have shown that expression of the S100B-EGFP transgene is initiated in Pax6+ precursors present in the RL at E12.5, which is after the bulk of DCN neuron precursors have left the RL 9 and reached the NTZ 8. The non-DCN nature of the RL Pax6+ precursors is substantiated by the existence of a clear boundary, at E13.5, between the EGFP-tagged RL and the EGFP-negative NTZ (Fig 5D). Such a discontinuity is not expected to occur when using a LacZ reporter gene driven by the Math1 locus 8.

Second, we think we can exclude UBC precursors as possible candidates for S100B-EGFP expression on the basis of their mode of migration that is different from the subpial mode of Pax6+ EGFP+ precursor migration within the EGL. Both populations of GPs and UBC precursors are known to share expression of the RL markers Pax6 and Math1. However, unlike UBC precursors which stream from the RL and posterior EGL into the developing white matter, and are therefore excluded from the anterior EGL 7, EGFP+ Pax6+ precursors invested the entire EGL with their leading process contacting the pial surface (Fig. 5A,C), a signature of migrating GPs 4446. Hence, we believe that transient expression of the S100B gene is a specific marker of the primary phase of GP genesis in the embryonic RL.

Finally, although we could not detect the S100B protein in the mouse RL, evidence exists for the presence of the S100B protein in the hamster EGL 32, and S100B mRNA in the murine cerebellar VZ including the RL 1234. Therefore, one interpretation for our results is that both the endogenous S100B gene and the transgene are expressed in the RL but the S100B protein cannot be detected because its cytoplasmic level is below detection limit, or because it is released in the intercellular space. The latter cannot be expected from the reporter EGFP protein since it is not a fusion of S100B and EGFP 11.

Before cerebellar midline fusion, the medial-most portion of the Cb neuroepithelium and abutting inferior tectal neuroepithelium, both co expressing EGFP and S100B at the highest level (Fig. 2), are also known territories of high En-1 expression 44. This designates En-1 as a potential inducer of S100B gene expression in the Cb. Beginning on E8.5 in the mouse, the Fgf-8 molecule produced in the isthmic orgnizer 47 induces expression of a series of genes including Wnt-1 and En-1 48. Wnt-1 maintains En-1 expression, which in turn, positively regulates Fgf-8, resulting in amplification of the Fgf-8 signal necessary for proper Cb specification and development 4950.

Interestingly, the earliest site of strong S100Bexpression in S100B-EGFP mice is the notochord beginning on E10.5 (Hachem and Legraverend, unpublished), and inhibition of En-1 expression by antisense targeting of early somite mouse embryos resulted in the loss of S100B expression in the notochord region subjacent to affected neural tube segments 51. Therefore, S100B may be a downstream component of the Wnt/En-1 regulatory cascade involved in the specification of both the notochord and Cb.

In the mouse embryonic Cb, radial glial cells and immature astrocytes express the intermediate filament vimentin 52, a protein with which S100B was shown to interact in vitro 53. S100B also binds to and activates in a Ca2+-dependent manner NDR-1 and NDR-2, two nuclear proteins that belong to an evolutionary conserved subfamily of serine/threonine protein kinases involved in the regulation of cell morphology 20. These types of interaction are consistent with S100B participating in vivo in the establishment and maintenance of a radial glial phenotype in the medial portion of the Cb where S100B is intracytoplasmic.

S100B may also be released from cerebellar rhombic lip EGFP+ Pax6+ precursors and promote the extension and branching of their neurites as illustrated in Fig. 6C. In 1985, Kligman and Marshak identified the extra cellular disulfide form of S100B as the molecule responsible for neurite extension of cultured chicken embryonic cortical neurons and named it "Neurite Extension Factor" 54. This NEF effect of S100B was later demonstrated on various types of neurons including cortical neurons 55, serotonergic mesencephalic neurons 56, dorsal root ganglion neurons 57 and spinal cord neurons 58, but the mechanism responsible for the NEF effect is unknown.

Another possible role exerted by the S100B protein released in the extracellular space is that of a trophic factor. Added in ovo at physiological concentrations, S100B was indeed capable of preventing the naturally occurring death of chicken motor neurons 59. The neurotrophic activity of S100B is thought to involve activation of NF-kappaB 60, binding to the receptor for advanced glycation end products (RAGE) and increased expression of the anti-apoptotic protein Bcl-2 21.

Transgenic S100B-EGFP mice 11, and S100B knockout mice 15 were housed under standard laboratory conditions in a 12-h light/dark cycle with access to food and water ad libitum. Experiments were performed according to the principles of laboratory animal care, following the guidelines approved by INSERM. Adult mice were allowed to mate overnight, and females were inspected for the presence of vaginal plugs the next morning (E0.5). Pregnant females were anesthetized with sodium pentobarbital and perfused through the ascending aorta with phosphate-buffered saline (PBS, pH 7.4) followed by 300 ml of fixative composed of 4% Para formaldehyde in 0.1 M phosphate buffer, pH 7.4. Embryos were quickly removed from the uteruses, anesthetized by hypothermia, and sacrificed by decapitation.

Embryos were cryoprotected by immersion in 30% sucrose in PBS for 12–24 h at 4°C, embedded in OCT compound (Tissue-Tek, Washington, DC) and frozen at -30°C. Tissue sections (15 μM) were obtained in a JUNG CM 300 cryostat (Leica), mounted onto poly-lysine-coated slides, and stored at -20°C. After several rinses in PBS, frozen sections were incubated in PBS containing 0.1% Triton X-100 and 10% goat serum for 1 h at room temperature (RT) and then incubated for 24 h at 4°C with primary antibodies followed by incubation with secondary antibodies for 1 h at RT. Primary antibodies were diluted in PBS containing 0.1% Triton X-100 and secondary antibodies were diluted in PBS containing 0.05% bovine serum albumin (BSA). After three rinses in PBS, sections were mounted in Mowiol (Calbiochem, La Jolla, CA) containing 2.5% 1,4-diazabicyclo- (2.2.2) octane (DABCO). The specificity of S100B immunolabeling was confirmed on sections of S100B null embryos (data not shown).

The following antibodies were used: rabbit polyclonal antibodies specific for S100B (1:1,000; Dako, Glostrup, Denmark, or Carpinteria, CA); Olig2 (1:6,000; obtained from Dr. Takebayashi, National Institute for Physiological Sciences, Okazaki, Japan); Pax6 (1:500; Berkeley Antibody Company, Richmond CA); Sox9 (1:500; Santa Cruz Biotechnology, Santa Cruz, CA); calbindin-1 (1:1000; Swant, Bellinzona, Switzerland). We used secondary donkey anti-rabbit IgG Cy3-F (ab0) 2 conjugates (1:1,000; Jackson ImmunoResearch, West Grove, PA). The specificity of S100B immunolabeling was confirmed on brain sections of S100B knockout embryos.

A Bio-Rad 1024 confocal laser scanning microscopic (CLSM) system equipped with an argon/krypton mixed gas laser was used for the analysis of adult and embryonic S100B-EGFP mice. Non-stacked CLSM images were obtained, representing optical sections with a depth of field of 0.5–3 μM. Two laser lines emitting at 488 nm and 568 nm were used for exciting EGFP and CY3-conjugated secondary antibodies, respectively. Four inputs were averaged to reduce the background noise, and green and red images were collected sequentially. Data acquisition and processing were controlled by the Laser sharp 1024 software and processing system. Unless otherwise stated, we marked EGFP+ Pax6- cells with asterisks and EGFP+ Pax6+ double positives with colored dots (green over red pax6+ nuclei, red over green EGFP+ cells). Pax6+ EGFP- cells were left unmarked. A cell was considered EGFP+ when both nucleus and cytoplasm were stained, or if a rim of green cytoplasm could be clearly assigned to a given nucleus. As a result, the number of double positives was most probably under evaluated. The pial and ventricular surfaces were outlined with white dashed lines; the limit between neuroepithelium and choroid plexus epithelium was marked with a blue dashed line; the choroid plexus epithelium of the fourth ventricle was outlined with a red dashed line.