CNTF/LIF/gp130 receptor complex signaling maintains a VZ precursor differentiation gradient in the developing ventral forebrain

The lateral ganglionic eminence (LGE) of the embryonic forebrain is the principle source of neurons that comprise the mature striatum(Deacon et al., 1994; Olsson et al., 1998; Olsson et al., 1995; Wichterle et al., 2001), and also contributes interneurons to the olfactory bulb and perhaps the neocortex(Corbin et al., 2001; Marin and Rubenstein, 2001). Like other regions of the telencephalon, two precursor cell populations reside within the LGE. The ventricular zone (VZ) forms first and VZ precursors undergo interkinetic nuclear migration during the cell cycle and mitotic division at the ventricular surface (Bhide,1996; Sheth and Bhide,1997; Smart,1976). At embryonic day (E) 11, the VZ gives rise to a secondary proliferative population that undergoes mitosis away from the ventricular surface and resides within the subventricular zone (SVZ)(Bhide, 1996; Halliday and Cepko, 1992; Sheth and Bhide, 1997; Smart, 1976). Precursors in the VZ are thought to produce the majority of the neurons generated from within the LGE, while SVZ precursors generate neurons(Anderson et al., 1997; Brazel et al., 2003; Sheth and Bhide, 1997) and glial cells (Brazel et al.,2003; Levison and Goldman,1993; Luskin and McDermott,1994). The presence of two germinal layers probably supports forebrain growth by overcoming congestion at the ventricular surface, allowing for expansion of the precursor pool and a prolonged period of proliferation by cells exiting the VZ (Smart,1972; Smart,1976). Indeed, in smaller structures such as the spinal cord,cells are derived from only a VZ during development(Smart, 1972; Smart, 1976). The signals responsible for the formation and maintenance of these two distinct germinal layers are largely unknown.

VZ precursors of the LGE express the homeobox transcription factor GSH2(Hsieh-Li et al., 1995; Yun et al., 2003), while more committed precursors in the SVZ are distinguished by increased expression of the proneural basic helix-loop-helix transcription factor, MASH1 (Ascl1 -Mouse Genome Informatics) (Casarosa et al.,1999; Fode et al.,2000; Porteus et al.,1994; Torii et al.,1999) and the Distal-less related homeobox genes, DLX1 and DLX2(Anderson et al., 1997; Eisenstat et al., 1999; Porteus et al., 1994; Yun et al., 2002). It has been suggested that MASH1 modulates Notch signaling in the germinal zone by regulating expression of the Notch ligands Delta1/3, and in the absence MASH1 (and hence Notch signaling) VZ precursors take on SVZ characteristics (Casarosa et al.,1999; Yun et al.,2002). Furthermore, overexpression of a proneural gene functionally related to MASH1, neurogenin 2 (NGN2), has been demonstrated to promote the acquisition of a SVZ fate by cells born in the cortical VZ(Miyata et al., 2004). These studies strongly suggest major roles for Notch signaling and proneural gene activity in the establishment and/or maintenance of the VZ and SVZ germinal compartments in the telencephalon.

We have previously reported that ciliary neurotrophic factor (CNTF)promotes the self-renewal of embryonic forebrain epidermal growth factor(EGF)-responsive neural stem cells (NSCs) in vitro(Shimazaki et al., 2001) by increasing the expression of NOTCH1 and inhibiting the expression of MASH1(Chojnacki et al., 2003),potentially conferring a more VZ like phenotype. CNTF signaling is transduced by a heterotrimer composed of the CNTFRα/LIFRβ/gp130 receptor complex, whereas LIF signaling is mediated by the LIFRβ/gp130 heterodimer, both of which can activate the JAK-STAT signaling cascade(Taga and Kishimoto, 1997). Others have found that LIF and gp130-mediated signaling can regulate the self-renewal of cortical NSCs in vivo(Hatta et al., 2002). In the present study, we tested the hypothesis that CNTF/LIF/gp130 receptor signaling contributes to the formation and maintenance of the germinal layers in the LGE during early stages of development. Our results suggest CNTF/LIF/gp130 receptor signaling is both necessary and sufficient to promote the self-renewal/expansion of a subpopulation of VZ precursors, establishing a VZ precursor differentiation gradient that is required for the normal growth of the ventral forebrain.

CD-1 mouse stocks were maintained in the University of Calgary Animal Resources Center (Calgary, Alberta, Canada). Breeding and genotyping of LIFRβ mutant mice has previously been described(Shimazaki et al., 2001). Morning of vaginal plug was considered embryonic day 0.5 (E0.5). All mice were sacrificed by cervical dislocation.

The generation and differentiation of neurospheres derived from the E14 CNS was performed as previously described with minor modifications(Represa et al., 2001; Reynolds et al., 1992; Shimazaki et al., 2001). Briefly, the lateral and medial ganglionic eminences or cervical/thoracic regions of the spinal cord were dissected from E14 mouse embryos, mechanically dissociated and plated as primary cultures at a density of 100,000 cells/ml in the presence of FGF2 (20 ng/ml; R&D Systems, Minneapolis, MN) and heparin sulfate (2 μg/ml; Sigma, St Louis, MO). After 7 days in vitro (DIV), the resulting primary neurospheres were mechanically dissociated and plated as pass 1 cells at 50,000 cells/ml. Additional cytokines used were CNTF (20 ng/ml; a generous gift from Dr Robert Dunn, Montreal General Hospital Research Institute) and LIF (20 ng/ml; Chemicon, CA, USA). To dissociate pass 1 neurospheres, we used a 3-minute treatment of 3 ml trypsin-EDTA (Invitrogen,Canada) at 37°C, followed by 5 ml trypsin inhibitor (Sigma). Cells were centrifuged at 600 rpm for 10 minutes and resuspended in 5 ml of trypsin inhibitor and mechanically dissociated with a fire-polished pipette. Cells were then suspended in 8 ml of basal media, centrifuged, resuspended in 2 ml of basal media and then used for experiments. For quantification of nestin+cells, pass 1 neurospheres were dissociated, plated at 200,000 cells/ml on poly-l-ornithine coated coverslips and fixed after 30 minutes for processing for immunocytochemistry. For quantification of β-tubulin III+ cells, pass 1 neurospheres were dissociated, plated at 200,000 cells/ml for 5 DIV and fixed for immunocytochemical analysis. For quantification of NSC self-renewal,dissociated pass 1 neurospheres were plated in 96-well plates at a density of 10,000 cells/ml and the number of secondary neurospheres was quantified after 10 DIV.

The forebrains of E12.5 Lifr^-/-,Lifr^+/- and Lifr^+/+ embryos were serially sectioned and collected onto two rounds of six slides with eight sections per slide and the slides were stained with Hoechst (Sigma). Slides were randomly sampled from each embryo and sections of the LGE and cortex were photographed beginning at the most rostral appearance of the LGE and extending caudally to the first appearance of the dorsal thalamus. The LGE was outlined and the area was calculated using Image J software (NIMH, MD, USA). The areas of the Lifr^-/- LGE and cortex were normalized to the averaged areas of Lifr^+/- and Lifr^+/+littermates for statistical analysis.

Pregnant dams received a single injection of bromodeoxyuridine on E12.5(BrdU; 120 mg/kg, ip; dissolved in 0.007% NaOH in phosphate buffer; Sigma) and were sacrificed thirty minutes post-injection. The brain and upper thoracic spinal cord was removed from E12.5 embryos and processed for immunohistochemistry as described below. Brains and spinal cords were sectioned serially onto two rounds of six slides with eight 14 μm sections per slide. Sections were pretreated with 1 M HCl for 30 minutes at 60°C to denature the DNA. Rat monoclonal anti-BrdU (1:50; Sera-Lab, Sussex, UK) and biotinylated donkey anti-rat (1:200; Jackson ImmunoResearch, West Grove, PA)with Streptavidin-Cy3 (1:2000; Jackson ImmunoResearch) were used for BrdU detection. The number of BrdU-expressing cells in the LGE or cortex was quantified on randomly sampled slides, beginning at the first appearance of the dorsal thalamus and extending rostrally for five sections for each animal. The same methodology was used for GSH2, MASH1 and pHH3 quantification.

Apoptosis was detected in tissue sections using the In Situ Cell Death Labeling Kit (Roche; Indianapolis, IN) according to the manufacturer's instructions with the following modifications. Sections were incubated in 100μl Proteinase K (diluted 1:1000 from a stock solution of 20 μg/μl in 10 mM Tris-HCl pH 8.0) for 6 minutes at room temperature and then washed extensively in PBS. The slides were incubated for 2 hours at 37°C in the presence of the TUNEL reaction mixture supplied by the kit, were washed in PBS, and then incubated overnight in the presence of mouse anti-FITC (1:400;Sigma). Detection was performed using biotinylated goat anti-mouse (1:200;Jackson ImmunoResearch) and Streptavidin-FITC (1:1000; Jackson ImmunoResearch). We quantified only those cells that were positive on the FITC channel and therefore not autoflourescent. Quantification was carried out in the same manner as for BrdU and the number of TUNEL-positive cells was normalized to the calculated area of the LGE, as described above.

Explant cultures were performed as described previously(Kingsbury et al., 2003) with modifications. Each cerebral hemisphere was dissected away from the developing E14.5 CNS in PBS. Next, the choroid plexus and regions of the septum and medial cortex were carefully dissected away to allow the culture media access to the ganglionic eminences. One hemisphere from each brain was placed in basal media (Opti-MEM I; Gibco, NY, USA) and the other was placed in basal media containing 50 or 100 ng/ml CNTF (either concentration was effective) in six-well plates (Nunc). The explants were placed on a rocker in a tissue culture incubator at 37°C, 5% CO₂ in air, for 18.5 hours, fixed in 4% paraformaldehyde at 4°C for 2 hours and cryoprotected in 25% sucrose overnight. Tissue was cryosectioned and processed for immunohistochemistry and quantification as described above for embryonic tissue. Explants that appeared damaged or unhealthy were excluded from analysis.

The choroid plexus was dissected from the lateral ventricles of 10-12 E14.5 CD-1 embryos and pooled for each round of RNA extraction. RNA was isolated using the RNeasy Protect Mini Kit (Qiagen, ON, Canada) and reverse transcribed using the Superscript III First-Strand Synthesis Kit (Invitrogen, CA, USA)with random hexamer primers, all according to the manufacturers instructions. PCR analysis of the resulting cDNA was performed to determine the expression of LIF using methods previously described(Bauer et al., 2003). CNTF cDNA was amplified by PCR for 35 cycles of 95°C for 45 seconds, 55°C for 45 seconds and 72°C for 45 seconds using 5′-ATGGCTTTCGCAGAGCAA-3′as sense primer and 5′-CTACATTTGCTTGGCCC-3′ as the antisense primer (Subang et al., 1997),which amplified a 596 bp fragment. CNTF and LIF RT-PCR products were sequenced by the University of Calgary DNA Sequencing Laboratory and checked for accurate identification by a BLAST search of the mouse genome.

The primary antibodies used in this study were as follows: rabbit anti-NOTCH1, anti-FGFR1, anti-FGFR2, anti-LIFRβ, anti-gp130, anti-CNTF and anti-LIF (1:50, Santa Cruz Biotechnology); goat anti-CNTFRα (1:50,Santa Cruz Biotechnology); mouse anti-Ki67 (1:100; NovoCastra Laboratories,Newcastle, UK); mouse anti-β-III tubulin (1:100; Sigma); rabbit anti-GSH2(generous gift from Dr Kenneth Campbell, Children's Hospital Research Foundation, Cincinnati); mouse anti-MASH1 (generous gift from Dr David Anderson, Caltech); rabbit anti-DLX1 and rabbit anti-DLX2 (1:100; generous gifts from Dr David Eisenstat; University of Manitoba); mouse anti-GAD65(1:100; BD Pharmingen); and rabbit anti-phospho Histone H3 (pHH3; 1:100;Upstate Biotechnologies).

For immunohistochemical analysis on embryonic tissue sections, timed pregnant mice were killed, the brain or upper thoracic spinal cord was removed from the embryos, fixed in 4% paraformaldehyde for 2 hours at 4°C and then cryoprotected by 10% sucrose and then 25% sucrose solutions overnight. All tissue was embedded in Tissue Tek OCT compound (Sakura Finetek, Torrance, CA)and cryosectioned at 14 μm. For receptor staining, tissue was post-fixed in methanol at -20°C for 8 minutes. Primary antibodies were followed by incubation with fluorescein- or rhodamine-conjugated secondary antibodies against mouse, rabbit or goat IgG, or by using biotin-conjugated secondary antibodies (all 1:200) followed by Streptavidin-CY3 (1:2000; Jackson ImmunoResearch). For nuclear counterstaining, Hoechst was used (1:100;Sigma).

Values are mean±s.e.m. Statistical significance between groups was assessed using ANOVA and the Tukey Honest Significant Difference Test, or paired t-test where noted.

To determine whether CNTF/LIF/gp130 signaling acts upon VZ and/or SVZ LGE precursors during development, we analyzed the expression pattern of the receptors in the LGE. At early stages of forebrain development, the majority of VZ precursor cells proliferate in response to fibroblast growth factor(FGF), rather than EGF (Martens et al.,2000; Raballo et al.,2000; Tropepe et al.,1999; Vaccarino et al.,1999). As expected, we observed both FGFR1 and FGFR2 immunoreactivity within the neuroepithelium of the ventral forebrain as early as E9.5 (Fig. 1A,B). At this stage, CNTFRα expression is absent(Fig. 1C); however, the expression of LIFRβ is robust throughout the ventral forebrain neuroepithelium (Fig. 1D). By E11.5, with the first appearance of both a VZ and SVZ(Smart, 1976), FGFR1, FGFR2,CNTFRα and LIFRβ are all expressed within the germinal zone of the LGE (Fig. 1E-H).

To determine whether the CNTF/LIF/gp130 and FGF signaling pathways might be regulating VZ precursors and/or SVZ precursors, we performed a series of double labeling experiments with Ki67, a marker specific for dividing cells,at E14.5 when both the VZ and SVZ are prominent(Smart, 1976). Double labeling for FGFR2 and Ki67 at E14.5 revealed co-expression in dividing VZ precursors,but not by those in the SVZ (Fig. 1I,J). Similarly, LIFRβ is expressed by VZ precursors, but not by Ki67-positive cells within the SVZ(Fig. 1K,L). These results suggest that CNTF/LIF/gp130 receptor signaling might act in coordination with FGF signaling to maintain precursors within the VZ, rather than those in the SVZ of the developing LGE.

The relationship between CNTF/LIF/gp130-mediated signaling and FGF signaling with respect to NSC regulation is not known. To assess whether CNTF or LIF signaling modulates FGF-responsive NSC proliferation, NSCs residing within the E14.5 ganglionic eminences were isolated and grown as neurospheres in FGF2. These neurospheres were dissociated, reseeded at 50,000 cells/ml in FGF2 and the presence or absence of either CNTF or LIF, and the total number of cells generated in each growth condition was assessed after 7 DIV. Remarkably, the number of cells generated by ventral forebrain NSCs in the presence of FGF-2 (Fig. 1M;2.9±0.3×10⁶) doubled in response to the addition of either CNTF (6.0±0.8×10⁶; P=0.009; n=5) or LIF (5.4±0.5×10⁶; P=0.03; n=5). To determine whether or not this increase might be due to an increase in the total number of neurospheres generated, primary FGF2 neurospheres were dissociated and replated in 96-well plates in FGF2 and the presence or absence of CNTF or LIF. After 10 DIV, we found no difference in the number of neurospheres generated in the presence of CNTF or LIF relative to FGF2 alone (Fig. 1N);however, the size of the neurospheres increased in CNTF- and LIF-treated conditions (Fig. 1O). Quantification revealed over a two-fold increase in the number of large neurospheres (>200 μm in diameter) in the CNTF and LIF growth conditions compared with FGF2 alone (Fig. 1P; CNTF, P=0.0006; LIF, P=0.005; n=3).

One possible explanation for the increased size of the CNTF or LIF treated neurospheres is that CNTF/LIF/gp130 signaling promotes NSC self-renewal and expansion, as we suggested previously for EGF-responsive NSCs(Shimazaki et al., 2001). To determine the number of precursors within the neurospheres cultured in the absence or presence of CNTF or LIF, we counted (in each growth condition) the percentage of cells that expressed nestin, a marker of undifferentiated cells(Frederiksen and McKay, 1988). In the CNTF/LIF-treated cultures, we observed a ∼50% increase in the percentage of cells expressing nestin relative to cultures generated in FGF2 alone (Fig. 1Q; CNTF P=0.03; LIF P=0.04; n=3). Additionally, when neurospheres generated in either FGF2, FGF2+CNTF or FGF2+LIF were dissociated and replated in 96-well plates in the presence of FGF2, the number of secondary neurospheres generated after 10 DIV doubled in the CNTF- and LIF-treated cultures (Fig. 1R;CNTF, P=0.02; LIF, P=0.02; n=3). These results suggest that CNTF/LIF/gp130 signaling promotes the self-renewal/expansion of FGF-responsive NSCs residing within the VZ.

Unlike the forebrain, the developing spinal cord has only a single germinal layer, called the VZ. We therefore sought to determine whether CNTF/LIF/gp130 signaling might play a different role in the regulation of spinal cord NSCs compared with forebrain NSCs. At E11.5, FGFR2 expression is clearly present in the spinal cord VZ (Fig. 2A). However, CNTFRα expression is primarily in the mantle zone(Fig. 2B), the region of neuronal maturation as shown by β-tubulin III expression(Fig. 2C). By E14.5, the spinal cord germinal zone is reduced in size, as shown by FGFR2 expression(Fig. 2D). The expression of CNTFRα (Fig. 2E)continued to predominate in the mantle zone(Fig. 2F), although low levels of expression are evident in the VZ. These results suggest that CNTF/LIF/gp130-mediated signaling might primarily influence cells of the neuronal lineage in the mantle zone rather than NSCs within the VZ of the spinal cord.

To examine the actions of CNTF/LIF upon the self-renewal/expansion of spinal cord NSCs, pass 1 neurospheres were generated from the spinal cord of E14.5 embryos in FGF2 and the absence or presence of CNTF or LIF. These neurospheres were dissociated and replated in 96-well plates in the presence of FGF2 alone, as was done to examine the self-renewal/expansion of ventral forebrain NSCs. Remarkably, the addition of either CNTF or LIF significantly inhibited the self-renewal of spinal cord NSCs, resulting in a 50% reduction in the number of secondary neurospheres formed(Fig. 2G; CNTF, P=0.003; LIF, P=0.007; n=4). This was the opposite effect to that observed with ventral forebrain NSCs.

Previous work has demonstrated a role for CNTFRα and LIFRβsignaling in regulating the differentiation/survival of spinal motoneurons(Li et al., 1995; Murphy et al., 1997; Nakashima et al., 1999; Richards et al., 1992). Given that we observed the majority of CNTFRα expression in the mantle zone,we asked whether the addition of CNTF/LIF to spinal cord neurosphere cultures affected neurogenesis by quantifying the number of β-tubulin III+neurons. The addition of either CNTF (^**P=0.007; n=4) or LIF (^**P=0.0001; n=4) to E14.5 pass 1 spinal cord neurosphere cultures resulted in a doubling of the number of neurons generated by spinal cord NSCs, compared with FGF2 alone(Fig. 2H). By contrast, there was no change in the number of neurons generated by E14.5 ventral forebrain neurosphere cultures treated with either CNTF or LIF(Fig. 2I; n=3).

To determine whether CNTF/LIF/gp130 signaling was required in vivo for the normal growth of the forebrain during development, we analyzed Lifrnull mutant mice, in which both CNTF and LIF signaling are disrupted. The forebrains of E12.5 Lifr^+/+, Lifr^+/-and Lifr^-/- littermate embryos were serially sectioned and the gross morphologies of the forebrains were compared. No obvious difference in forebrain growth was observed between the Lifr^+/+ and Lifr^+/- littermates (data not shown). On the other hand,although an obvious difference in overall brain size was not noted, a histological comparison of sections counterstained with the nuclear marker Hoechst from rostral to caudal between Lifr^+/+/Lifr^+/- and Lifr^-/- mice revealed gross morphological differences that were suggestive of impaired growth in the mutant forebrains(Fig. 3). At rostral levels,the most obvious difference was a decrease in the size of the LGE in the Lifr^-/- mice (Fig. 3A,B). At more caudal regions of the forebrain, we observed that the cortex appeared thinner compared with Lifr^+/+/Lifr^+/- controls, and the LGE,MGE and caudal ganglionic eminence were also reduced in size(Fig. 3C-H). This phenotype was partially penetrant with 56% (n=9) of the Lifr^-/-embryos displaying gross morphological changes relative to littermate Lifr^+/+/Lifr^+/- controls.

Further analysis was focused on the LGE at E12.5, a stage when the LGE is clearly distinguishable from the MGE and VZ and SVZ germinal layers are present (Smart, 1976). Measurements of the LGE (see Materials and methods) in Lifr^+/+/Lifr^+/- versus Lifr^-/- embryos revealed a 22% reduction in the size of the LGE in affected Lifr^-/- embryos (paired t-test ^**P=0.008; n=5; Fig. 3I,J). Immunohistochemical analysis for β-tubulin III demonstrated a concurrent decrease in the size of the mantle zone (Fig. 3K,L),which suggested either a decrease in overall growth or an increase in cell death. The number of TUNEL+ cells in the LGE was quantified and normalized to the area. No difference in the number of cells undergoing cell death between the Lifr^-/- (0.5±0.1 cells/unit area) and Lifr^+/+/Lifr^+/- embryos(0.5±0.1 cells/unit area; n=4) was observed. To determine whether the defect might instead be related to fewer precursor cells in the germinal zone, proliferating precursors were labeled by injecting dams with BrdU. The number of BrdU+ cells in the VZ/SVZ of the LGE of Lifr^-/- embryos (608±15; Fig. 3O) was reduced by 30%relative to Lifr^+/+/Lifr^+/- embryos(865±41; P=0.009; paired t-test; n=5; Fig. 3P). These observations were not limited to the LGE and nearly identical defects were observed in the developing cortex (see Fig. S1A-H in the supplementary material). By contrast,however, the number of BrdU positive cells in the E12.5 spinal cord VZ was not significantly decreased in Lifr^-/- relative to Lifr^+/+/Lifr^+/- embryos, and no obvious morphological changes related to growth were observed (see Fig. S2A,B in the supplementary material). These initial in vivo analyses suggested a region specific role for CNTF/LIF/gp130 signaling in regulating the self-renewal/expansion of telencephalic precursors and justified further investigations into a potential role for this pathway in establishing the VZ and SVZ precursor populations of the LGE.

To investigate the role of CNTF/LIF/gp130 signaling in the formation and maintenance of the ventral forebrain germinal layers, it was necessary to establish criteria to distinguish undifferentiated VZ precursor cells from more differentiated precursors. Previous studies have suggested that the transcription factors GSH2, MASH1 and DLX2 might be useful as markers of VZ precursor differentiation (Eisenstat et al., 1999; Porteus et al.,1994; Yun et al.,2002; Yun et al.,2003), prompting us to examine their expression in more detail.

At E8.5, neither GSH2 nor MASH1 were expressed in the VZ (data not shown),in agreement with previous studies (Yun et al., 2002), and we have designated these cells rudimentary precursor cells (RPs). At E9.5, when the germinal zone is composed of a VZ only, a large number of precursor cells expressed GSH2 in the LGE, and we have designated these as P1 precursor cells(Fig. 4A). A subpopulation of VZ precursors also expressed MASH1 (Fig. 4B), and all the MASH1+ cells co-expressed GSH2(Fig. 4C). Very few DLX2+ cells were detected at the pial surface of the E9.5 LGE VZ(Fig. 4D), and these cells co-expressed MASH1 (Fig. 4E,F). The DLX2+ subpopulation was presumed to express GSH2, as all MASH1+ cells at this stage appeared to express GSH2 (the triple labeling experiment was not possible). This analysis, at a stage when the SVZ is absent(Smart, 1976), suggests that the LGE VZ is composed of P1, P2 and P3 precursors, as depicted in Fig. 4P.

By E12.5, the LGE germinal zone is composed of both VZ and SVZ precursors(Sheth and Bhide, 1997; Smart, 1976). GSH2 expression continued to be heavily expressed by VZ precursors(Fig. 4G), whereas MASH1+ cells were organized in a gradient, with fewer precursors expressing low levels in the VZ, to more precursors expressing higher levels in the SVZ(Fig. 4H). GSH2/MASH1 double-labeling revealed a VZ to SVZ molecular gradient from GSH2+ only precursors (P1 precursors) at the ventricular surface, to GSH2+MASH1+precursors and then MASH1+ only precursors in SVZ(Fig. 4I). Analysis of DLX2 and MASH1 expression revealed a molecular gradient of low DLX2 and MASH1 at the ventricular surface to increased numbers of MASH1+ precursors, MASH1+DLX2+precursors, and, finally, DLX2+ precursors at the superficial edge of the SVZ and in the mantle zone (Fig. 4J-L). This analysis expands on previous studies(Yun et al., 2002) and suggests that a gradient of LGE precursor differentiation exists that can be defined by GSH2, MASH1 and DLX2 expression, as summarized in the schematic in Fig. 4P.

To understand how CNTF, LIF and FGF signaling might regulate this differentiation gradient, we performed a series of double-labeling experiments. Analysis of the expression of CNTFRα and GSH2 revealed that a subpopulation of GSH2+ VZ precursors located close to the ventricular surface express CNTFRα (Fig. 4M; arrow indicates a double-labeled cell). LIFRβ, required for both CNTF and LIF signaling, is also expressed by VZ precursors close to the ventricular surface. Double labeling revealed that very few LIFRβ+cells co-express MASH1 (Fig. 4N; arrow indicates an example of a double-labeled cell). Similarly, only a small subpopulation of MASH1+ cells co-expressed FGFR2(Fig. 4O). These results suggest that CNTF, LIF and FGF signaling might act primarily to regulate the self-renewal/expansion of the VZ P1 precursor (GSH2+MASH1-) cell population,perhaps inhibiting progression to more differentiated precursor fates (P2-P5),as depicted in the schematic in Fig. 4P.

We hypothesized that impaired self-renewal/expansion of P1 precursor cells,owing to lost CNTF/LIF/gp130 signaling, results in the precocious differentiation of VZ precursors. To test this, we analyzed the number of GSH2+ VZ precursors in the LGE of E12.5 Lifr^-/- animals relative to Lifr^+/+/Lifr^+/- embryos. Lifr^-/- embryos had ∼30% fewer GSH2+ precursors(Fig. 5B; 911±23)relative to Lifr^+/+/Lifr^+/- controls(Fig. 5A; 1324±44;paired t-test ^**P=0.0006; n=4). When normalized to the decreased area of the LGE, the overall decrease in the Lifr^-/- GSH2+ population was 15% relative to Lifr^+/+/Lifr^+/- littermates (paired t-test ^*P=0.04). By contrast, the number of MASH1+ precursors within the germinal zone of the Lifr^-/-embryos (Fig. 5D;802±30) was increased by 37% compared with Lifr^+/+/Lifr^+/- embryos(Fig. 5C; 504±22; paired t-test ^**P=0.0007; n=3). Double-labeling of MASH1 and GSH2 demonstrated a dramatic increase in the MASH1+GSH2-cell population within VZ of the Lifr^-/- embryos(Fig. 5E,F). Analysis of DLX2 also revealed increased expression by VZ precursors within the Lifr^-/- LGE (Fig. 5G,H). Double labeling with MASH1 and DLX2 demonstrated that the majority of the DLX2+ cells within the VZ of the Lifr^-/-embryos co-expressed MASH1 (Fig. 5I,J).

To confirm that VZ precursors within the LGE of Lifr^-/-embryos precociously differentiate, we analyzed the expression of the transcription factor DLX1 (Fig. 5K,L), which is normally largely restricted to the SVZ(Eisenstat et al., 1999), as well as the enzyme GAD65 (Fig. 5M,N), which is expressed exclusively in the SVZ and mantle zone(Stuhmer et al., 2002). DLX1 expression was clearly upregulated within the VZ of the Lifr^-/- embryos (Fig. 5K,L). Furthermore, GAD65 expression, which normally was restricted to the SVZ and mantle zone (Fig. 5M), was ectopically expressed within the VZ of the Lifr^-/- embryos (Fig. 5N; arrow indicates ectopic VZ expression in Lifr^-/- embryos). Finally, we took advantage of the fact that VZ precursors undergo mitosis at the ventricular surface and SVZ precursors undergo non-surface mitosis. VZ and SVZ precursors were quantified using pHH3, which marks cells in m-phase. The number of surface pHH3+ cells(VZ precursors) was reduced by 40% in the LGE of the Lifr^-/- embryos compared with littermate Lifr^+/+/Lifr^+/- embryos(Fig. 5O,P; paired t-test ^**P=0.0005; n=4). However, the number of non-surface pHH3+ cells (SVZ precursors) was not significantly different between the Lifr^-/- and Lifr^+/+/Lifr^+/- littermates. Therefore, the ratio of SVZ precursors to VZ precursors was increased in the Lifr^-/- embryos (similar results were attained for the cortex as shown in see Fig. S1I,J in the supplementary material). These results suggest that in the absence of CNTF/LIF/gp130 signaling, VZ precursors downregulate GSH2 expression and precociously differentiate to a more committed cell fate.

To understand better how CNTF/LIF/gp130 signaling might operate to maintain VZ precursors, we attempted to determine the likely source of CNTF and LIF acting upon the VZ during forebrain development. Immunostaining for CNTF expression in the forebrain of E14.5 embryos revealed robust expression within the choroid plexus, as well as at the base of the choroid plexus(Fig. 6A,B; arrow and arrowheads indicate staining). Immunoreactive cells were also observed within the mantle zone of the LGE (Fig. 6B); however, it is unlikely that these cells would greatly influence precursors within the VZ. LIF expression was also robust in the choroid plexus (Fig. 6C,D). We further confirmed the expression of both CNTF and LIF in the E14.5 choroid plexus by RT-PCR (Fig. 6E). The choroid plexus generates the cerebrospinal fluid (CSF) that bathes the interior of the embryonic and adult brain, and it is known to secrete many different growth factors and cytokines(Chodobski and Szmydynger-Chodobska,2001; Dziegielewska et al.,2001; Gard et al.,2004; Speake et al.,2001; Sturrock,1979). Our results suggest that the embryonic choroid plexus also generates CNTF and LIF. The presence of these molecules in the CSF would place them in an ideal position to influence precursors at the ventricular surface and establish a gradient of VZ precursor differentiation, as depicted in the model in Fig. 6F.

In order to test the role of the choroid plexus, and the sufficiency of CNTF/LIF/gp130-mediated signaling to maintain the VZ differentiation gradient,we performed the following experiment (see schematic in Fig. 7A). Cerebral hemispheres were dissected from E14.5 embryos and the medial cortex, septum and choroid plexus were removed. The hemispheres were then cultured as explants in either basal media, thereby removing the CSF as a source of CNTF and LIF, or basal media containing CNTF for 18.5 hours. Explants were first analyzed for changes in the level of proliferation within the LGE by staining for pHH3. The number of proliferating cells in the LGE germinal zone of CNTF-treated explants(235±27) was increased by 79% relative to the number observed in the explants cultured in basal media alone(Fig. 7B-D ;131±19; paired t-test ^**P=0.009; n=5). Therefore, CNTF treatment was sufficient to increase the number of proliferating cells in germinal zone of the LGE.

We next sought to determine whether this increased number of proliferating cells in response to CNTF treatment might be related to maintenance of the graded differentiation of VZ precursors. The number of GSH2+ cells within the germinal zone of the CNTF-treated explants (1107±166) increased by 51%relative to the basal media-treated explants (729±81; paired t-test ^*P=0.03; n=4), though this population appeared diminished in both conditions relative to the normal E14 LGE (Fig. 7E-G). MASH1 was clearly expressed in a gradient from VZ to SVZ in the normal E14.5 LGE(Fig. 7H). However, in the basal media-treated explants, the expression gradient was completely disrupted and MASH1 was expressed by VZ precursors abutting the ventricular surface(Fig. 7I; n=5),similar to the Lifr^-/- phenotype. Remarkably, the addition of CNTF to the explant condition was sufficient to rescue this defect, as these explants clearly retained a VZ to SVZ gradient of MASH1 expression(Fig. 7J; n=5). Similarly, while DLX1 expression was present in a VZ to SVZ gradient in the normal E14.5 LGE (Fig. 7K),this expression pattern was disrupted in the basal media-treated explants in a manner similar to MASH1, whereby DLX+ cells were located near the ventricular surface (Fig. 7L). The gradient of DLX1 expression was also restored in the CNTF-treated explants(Fig. 7M).

Previously, we have reported that CNTF signaling can regulate the expression of NOTCH1 (Chojnacki et al.,2003) and a disruption of Notch signaling has been suggested to result in the precocious expression of SVZ characteristics by VZ precursors(Casarosa et al., 1999; Yun et al., 2002). Therefore,we examined the expression of the NOTCH1 receptor in our explant culture conditions. In the normal E14.5 LGE, NOTCH1 expression appeared in a gradient opposite to MASH1, being highest at the ventricular surface and decreasing in expression away from the surface (Fig. 7N). In the basal media explant condition NOTCH1 expression was sustained; however, the gradient was disrupted and replaced by a cluster of NOTCH1 immunoreactivity along the ventricular surface(Fig. 7O; n=5). The gradient was restored in the CNTF-treated explants(Fig. 7P; n=5). Double-labeling with NOTCH1 and MASH1 in the normal E14.5 LGE revealed that MASH1 expression was restricted to cells expressing low levels of NOTCH1 at superficial levels of the NOTCH1 gradient or by cells that did not express NOTCH1 in the SVZ (Fig. 7Q). Remarkably, in the basal media explants many of the clustered NOTCH1-expressing cells co-expressed MASH1(Fig. 7R; n=5), while CNTF treatment maintained the graded separation of these two cell populations(Fig. 7S; n=5).

In this study, we tested the hypothesis that CNTF/LIF/gp130 signaling plays a key role in the formation and maintenance of the unique laminar organization of the ventral forebrain germinal zone. Our results suggest that CNTF/LIF/gp130 receptor complex signaling regulates the self-renewal/expansion of a subpopulation of VZ precursors and that this process is both necessary and sufficient to maintain the gradient of VZ precursor differentiation in the LGE. These findings clarify how the ventral forebrain is able to establish the germinal layers that support the growth of the developing striatum. In addition, our findings reveal how the CNTF/LIF/gp130 receptor complex signaling may operate regionally to regulate precursor cells in the developing CNS.

The signals that form and maintain the germinal layers of the forebrain are poorly understood. Building on the work of Yun et al.(Yun et al., 2002), we have proposed that a gradient of VZ precursor differentiation can be defined by GSH2, MASH1 and DLX1/2 expression. On this basis, precursor differentiation in the LGE can be considered to occur in three major steps: (1) VZ precursors are initially specified by expression of GSH1 and GSH2, which inhibit dorsal cell fate characteristics by repressing Pax6(Corbin et al., 2000; Schuurmans and Guillemot,2002; Toresson et al.,2000; Yun et al.,2003; Yun et al.,2001); (2) Precursors become initially committed to the neuronal lineage by the expression of MASH1(Casarosa et al., 1999; Fode et al., 2000; Nieto et al., 2001; Torii et al., 1999; Yun et al., 2002); and (3)MASH1 initiates the expression of the Dlx genes(Fode et al., 2000), which induce glutamic acid decarboxylase expression and regulate the generation of later born neuronal populations (Anderson et al., 1997; Eisenstat et al., 1999; Stuhmer et al.,2002; Yun et al.,2002). Recent time-lapse studies of cortical precursors have demonstrated that the SVZ arises from asymmetric divisions of precursors in the VZ (Miyata et al., 2004; Noctor et al., 2004), and a similar relationship probably exists between the VZ and SVZ in the LGE(Halliday and Cepko, 1992). Therefore, the differentiation gradient we have described in the LGE(summarized in Fig. 4P) may represent differentiating daughter cells, asymmetrically born at the ventricular surface from P1 precursors, trafficking through the VZ out to the SVZ.

A role for Notch signaling and proneural gene expression in maintaining the separate VZ and SVZ layers has been suggested. NGN2 promotes the commitment of VZ daughter cells to the SVZ in the cortex(Miyata et al., 2004). It is likely that MASH1 plays a similar role in the ventral telencephalon, given its known role in promoting precursor commitment and that NGN2 is sufficient to rescue neurogenesis defects in the ventral telencephalon of MASH1 mutant mice(Parras et al., 2002). By contrast, NOTCH1 signaling has been suggested to play a role in maintaining the VZ population of the ventral forebrain(Casarosa et al., 1999; Yun et al., 2002). We have previously demonstrated that CNTF/LIF/gp130 signaling supports embryonic NSC self-renewal in vitro (Shimazaki et al.,2001) by promoting NOTCH1 and repressing MASH1 expression(Chojnacki et al., 2003). Our present results suggest that FGF-responsive P1 precursors located at the ventricular surface are the primary targets of CNTF/LIF/gp130 signaling in the developing LGE. First, components of the CNTF/LIF/gp130 signaling pathway are expressed primarily by GSH2+MASH1-precursors at the ventricular surface. Second, this pathway promotes FGF-responsive NSC self-renewal/expansion in vitro and promotes the P1 precursor fate in forebrain explant cultures. Finally, in the absence of this signaling pathway VZ precursors precociously differentiate and acquire characteristics of more differentiated cells at the expense of the GSH2+ cell population. CNTF/LIF/gp130 signaling thus appears to maintain an undifferentiated P1 precursor population in the VZ, which contributes to the normal formation of two germinal layers having distinctly different degrees of cellular commitment.

It is reasonable to suggest that CNTF/LIF/gp130 signaling may be acting upstream of NOTCH1, as we have suggested previously(Chojnacki et al., 2003), or in coordination with NOTCH1 signaling(Kamakura et al., 2004) to maintain the P1 precursor population. Our explant culture experiments reveal that CNTF/LIF/gp130 signaling maintains NOTCH1- and MASH1-expressing cells largely as separate populations. It may be that in the absence of this signaling pathway fewer NOTCH1 receptors are expressed per P1 precursor cell,resulting in decreased Notch signaling and the precocious expression of MASH1 and other more differentiated cellular characteristics. Alternatively, the activation of STAT3 by CNTF/LIF/gp130 signaling may mediate important crosstalk between the Notch and STAT signaling pathways [as recently suggested by Kamakura et al. (Kamakura et al.,2004)], which is required to maintain the undifferentiated state of P1 precursors.

Our results suggest that the primary source of CNTF and LIF acting upon VZ precursors may be the choroid plexus, a secretory structure that gives rise to the CSF that bathes the interior of the developing brain(Dziegielewska et al., 2001; Speake et al., 2001). The choroid plexus is known to secrete many different growth factors and cytokines into the CSF for delivery to the brain tissue(Chodobski and Szmydynger-Chodobska,2001), appears as early as E11 in the forebrain(Sturrock, 1979), and is believed to produce CSF throughout development(Dziegielewska et al., 2001). The CSF of the developing embryonic forebrain is ideally positioned to influence precursor cells within the germinal zone such that cells closest to ventricular surface would be closest to the source of CNTF/LIF, while those positioned farther away from this source would be in an environment increasingly permissive for differentiation (see Fig. 6F). In support of this model, our results demonstrate that the removal of both the choroid plexus and the CSF in the basal media explant condition results in the disruption of the VZ precursor differentiation gradient. This defect was entirely rescued by the addition of CNTF to the media bathing the explant and concurrently increased the number of proliferating cells within the germinal zone.

It has been suggested that the adult choroid plexus does not express CNTF and LIF (Gard et al., 2004),which may be significant given that only the SVZ (and not the VZ) is maintained into adulthood. Perhaps the production of CNTF and LIF by the choroid plexus is required to maintain levels of these molecules sufficient for the formation of both the VZ and SVZ germinal layers. This is an area for future investigation that may be important for strategies aimed at the intrinsic repair of the forebrain.

Virtually all previously identified pathways that promote NSC self-renewal/expansion are known to function in both the forebrain and spinal cord. For example, FGF signaling promotes NSC self-renewal/expansion in both the forebrain (Martens et al.,2000; Raballo et al.,2000; Tropepe et al.,1999; Vaccarino et al.,1999) and spinal cord (Del Corral and Storey, 2004; Represa et al., 2001). Similarly, Notch signaling promotes NSC self-renewal/expansion in both forebrain (Chojnacki et al.,2003; Hitoshi et al.,2002; Schuurmans and Guillemot, 2002; Yun et al.,2002) and spinal cord regions(Appel et al., 2001; le Roux et al., 2003; Lindsell et al., 1996), as do the membrane-associated proteins Numb and Numblike(Petersen et al., 2002; Petersen et al., 2004). Our review of the literature suggests that a specific signaling pathway that promotes VZ precursor self-renewal/expansion within the forebrain, but not the spinal cord, has not been previously identified.

We found that the components of the CNTF/LIF/gp130 receptor complex were robustly expressed by VZ precursors in the developing LGE, contrasting with CNTFRα and LIFRβ expression in the spinal cord(Fig. 2) (C.G. and S.W.,unpublished), which was largely restricted to the mantle zone. A similar pattern of CNTFRα expression was previously observed in the developing rat spinal cord (MacLennan et al.,1996). This suggests that CNTF/LIF/gp130 signaling might have a different role in the spinal cord compared with the forebrain. Indeed, CNTF and LIF inhibited spinal cord NSC self-renewal/expansion in vitro, instead promoting the generation/survival of neurons from FGF-responsive spinal cord NSCs. Furthermore, we observed no impairment in the growth or number of proliferating precursors in the developing spinal cord of Lifr^-/- mice and the addition of CNTF to spinal cord explant cultures decreased the number of proliferating precursors in the spinal cord VZ (see Fig. S3A,B in the supplementary material). Taken together,these findings suggest that CNTF/LIF/gp130 receptor signaling specifically contributes to growth of the forebrain (but not the spinal cord) during development, by contributing to the formation and maintenance of the ventral forebrain VZ.

This work was supported by the Canadian Institutes of Health Research(CIHR). We thank Rozina Hassam and Dorothea Livingstone for excellent technical assistance, and Andrew Chojnacki, Kathryn Markham and Carol Schuurmans for critical review of an earlier version of this manuscript. C.G. is recipient of an Alberta Heritage Foundation for Medical Research (AHFMR)studentship. S.W. is an AHFMR Scientist.