GENERATION OF CAJAL-RETZIUS NEURONS IN MOUSE FOREBRAIN IS REGULATED BY TRANSFORMING GROWTH FACTOR β-FOX SIGNALING PATHWAYS

Abstract

The generation of Cajal-Retizus (CR) neurons is restricted to discrete sites in the telencephalon. Most of these sites do not express Foxg1, a transcription factor that inhibits transforming growth factor (TGF) β dependent up-regulation of p21. We tested the hypothesis that TGFβ signaling triggers CR neuronogenesis in Foxg1-deficient zones through p21 induction. In Foxg1^+/+ mice, p21 (a) was expressed in select cycling cells in CR neuron-producing areas and (b) was co-localized in newly generated CR neurons. Zones of CR neuronal production and p21 expression were expanded in the forebrains of Foxg1^Cre/Cre mice. Manipulation of TGFβ signaling in explants from cortical hems of wild-type mice altered p21 expression and the production of CR neurons. Furthermore, despite continued TGFβ activity, p21 immunoreactivity diminished in CR neurons with distance from their generation site. This implicated a second pathway controlling p21 expression. We provide evidence that Foxo3a, which has been shown to translocate into the nucleus to act as a transcriptional co-activator of TGFβ-dependent upregulation of p21, is strategically expressed to be involved in controlling p21 expression in CR neurons. Specifically, Foxo3a was nuclear in p21+/reelin+ cells in sites of CR neuronal generation, however, nuclear Foxo3a immunoreactivity was absent in p21−/reelin+ cells distal from sites of CR neuronogenesis. Thus, TGFβ and Foxo3a may work in concert to regulate expression of p21 during CR neuronal generation.

INTRODUCTION

Cortical development relies upon instructive cues that guide young neurons as they migrate to their eventual residences. One such cue is reelin, a large extracellular glycoprotein secreted by Cajal-Retzius (CR) neurons (Derer and Derer, 1991; D’Arcangelo et al., 1995; Hirotsune et al., 1995; Ogawa et al., 1995; Meyer and Wahle, 1999; Soda et al., 2003). In the absence of reelin, the normal inside-to-outside growth of cortex does not occur (Caviness, 1982).

CR neurons are among the first forebrain neurons generated. In mice, cortical CR neurons are derived between gestational day (G) 10.5 and G13.5 from specific regions of the telencephalic neuroepithelium (Hevner et al., 2003; Takiguchi-Hayashi et al., 2004; Bielle et al., 2005). CR neuronal production is restricted by the forkhead transcription factor Foxg1. In Foxg1-null mice, the medial cortical regions of CR neuronal generation are expanded at the expense of lateral neocortex, leading to an over-production of CR neurons (Hanashima et al., 2004; Martynoga et al., 2005; Muzio and Mallamaci, 2005; Zhao et al., 2006). Furthermore, Foxg1 expression is largely absent in regions of CR neuronal production (Tole and Patterson, 1995; Martynoga et al., 2005).

Foxg1 is a potent inhibitor of transforming growth factor (TGF) β-regulated signaling (Dou et al., 2000; Rodriguez et al., 2001). It does so by inhibiting TGFβ-dependent transcription of the cyclin-dependent kinase inhibitor (CKI) p21, a cell cycle protein that can induce cell cycle exit in neural precursors (Seoane et al., 2004; Siegenthaler and Miller, 2005). This is underscored by evidence that 12.5-day-old Foxg1-null mouse fetuses, with their over-production of CR neurons, have a greater number of p21-positive (p21+) cortical cells than age-matched wild-type (Foxg1^+/+) fetuses (Seoane et al., 2004).

Studies of non-neural cells show that another forkhead transcription factor, Foxo3a, can interact with downstream TGFβ signaling proteins to facilitate p21 transcription (Seoane et al., 2004). Foxo3a translocates to the nucleus and complexes with Smad3 and Smad4 to drive the transcription of p21. Nuclear translocation of Foxo3a is negatively regulated by the pathway initiated by insulin growth factor (IGF) 1 and transduced by phosphotidylinositol 3-kinase (PI3-K) and Akt (Fukunaga et al., 2005). Activated Akt phosphorylates Foxo3a and triggers its removal from the nucleus. This Foxo3a pathway has been implicated in ischemia-induced cell death in hippocampal neurons (Brunet et al., 1999). Growth promoting and anti-thanatopic factors, e.g., IGF1, bar Foxo3a from the nucleus, hence, preventing the Foxo3a-regulated transcription of cell death genes.

The observations that (a) Foxg1 is absent from regions of CR cell production and (b) pro-differentiation signals initiated by TGFβ (i.e., stimulation of p21 expression) are inhibited by Foxg1 underlie the hypothesis that TGFβ signaling is an inductive cue for CR neuronogenesis. A corollary is that nuclear shuttling of Foxo3a is also involved in regulating p21 expression in CR neurons. Thus, we examined (a) p21 and TGFβ signaling protein expression during CR neuronal generation in Foxg1 wild-type (Foxg1^+/+) and null (Foxg1^Cre/Cre) (b) the requirement of TGFβ signaling in one generation zone, the cortical hem and (c) the localization of Foxo3a in areas of CR neuron production.

MATERIALS AND METHODS

Tissue collection

To generate Foxg1^Cre/Cre mice, C57BL/6 Foxg1^+/Cre males and females (a generous gift from Pat Levitt, Vanderbilt University) were mated overnight. The first day the sperm-positive plug was observed was designated as G0.5. On G12.5 or G13.5, the dams were sacrificed and the fetal brains were harvested. Foxg1^Cre/Cre fetuses were distinguished from heterozygous littermates by the brain and eye abnormalities (Hebert and McConnell, 2000).

Tissue was collected for genotyping. The genotypes of breeding adult and fetal mice were determined using primers designed to amplify both the Foxg1 wild-type and null allele. Three primers were 5′-GCC GCC CCC CGA CGC CTG GGT GAT G-3′, 5′-TGG TGG TGG TGA TGA TGG TGA TGC TGG-3′, and 5′-ATA ATC GCG AAC ATC TTC AGG TTC TGC GGG-3′ (Muzio and Mallamaci, 2005).

After the tissue was harvested for genotyping, anesthetized fetuses were prepared in one of three ways. (1) Some fetuses were euthanized by immersion fixation overnight at 4°C in 4.0% paraformaldehyde. These brains were removed, cryoprotected in sucrose buffers, frozen, and cut into a series of 12 μm sections. (2) The brains of other fetuses were removed and prepared for organotypic cultures. (3) Samples of unfixed brain tissue or slices were used in immunoblotting studies to determine the expression of Foxg1 and phosphorylated Smad (pSmad). The procedures used in each of these preparations are described below.

Organotypic slice cultures

Cortical slices were obtained from the brains of wild-type, 13.5-day-old fetuses. Brains were collected in Krebs buffer (126 mM NaCl, 2.5 mM KCl, 1.2 mM NaH₂PO₄·H₂0, 1.2 mM MgCl₂,2.5 mM CaCl₂, 11 mM glucose, 25 mM NaHCO₃, 10 mM HEPES) and cut into 300 μm coronal sections using a MacIlwain Tissue Chopper (Mickle Lab Engineering, Gomshell UK).

Slices were cultured on filter inserts with 0.40 μm pores (Millipore, Bedford MA) in a medium composed of Neurobasal Medium (Invitrogen, Carlsbad CA), 2.0% B-27 supplement (Invitrogen), 2.0 mM glutamine (Invitrogen), 100 mM dextrose, and 100 μM penicillin/streptomycin (Invitrogen). The cultures were incubated at 37°C with 6.0% CO₂. Following two hours in culture, some slices were treated with TGFβ1 (40 ng/ml; Sigma, St. Louis MO), insulin-like growth factor-1 (300 ng/ml; IGF-1; Sigma), SB431542, a blocker TGFβ receptor activity (100 μM; Tocris Bioscience, Ellisville MO), or LY-294002, an inhibitor of phosphotidylinositol 3-kinase activity (PI3-K; 60 μM; Sigma). Control slices were treated with an equal volume of 0.40% dimethylsulfoxide, the vehicle for SB431542 and LY-294002. Slices were incubated in the various treatment conditions for 18 hr and then fixed in 4.0% paraformaldehyde for 30 min. The fixed slices were processed for cryosectioning, frozen, and cut into 12 μm sections.

Immunohistochemistry

Immunolabeling of sections, be they from whole brains or slices, were processed by the same procedure. Prior to incubation with a primary antibody(ies), all sections were incubated in 3.0% H₂O₂ for 5 min then steamed in 0.010 M citric acid for 15 min. After steaming, the sections were cooled to room temperature in PBS. Non-specific immunoreactivity was blocked with a wash in a solution of 1.0% bovine serum albumin and 0.75% Triton-X in PBS for 45 min.

After the blocking step, each section was immunoreacted with a primary antibody(ies) (Table 1). For multiple labeling including p21, the Tyramide System Amplification (TSA) Kit #10 (Molecular Probes, Eugene OR) was used as per manufacturer’s instructions. Immunolabeling for pSmad2 relied on TSA Kit #12. The appropriate fluorescein-conjuated secondary antibodies (1:200; Jackson Immunologicals) were used to tag the primary antibodies.

Table 1.

Primary antibodies.

Primary antibody	Source	Catalog Number	Dilution	Host
actin	Sigma, St. Louis MO	A5316	1:5000 #	mouse
Foxg1	Yoshiki Sasai, RIKEN Institute, Kobe, Japan	- - -	1:500 *	rabbit
Foxo3a	Sigma	F2178	1:200 *	rabbit
Ki-67	LabVision, Fremont CA	RM-9106-S	1:200 *	rabbit
PCNA	BD Biosciences, San Diego CA	610664	1:100 *	mouse
p21	BD Biosciences	556431	1:2000 *	mouse
p73	LabVision	MS-762-PO	1:200 *	mouse
reelin	Chemicon, Temecula CA	MAB5364	1:300 *	mouse
Smad2/3	Cell Signaling Technologies, Beverly MA	3103S	1:800 #	mouse
pSmad2	Cell Signaling	3101L	1:500 * 1:100 #	rabbit
TGFβrI	Santa Cruz Biotechology, Santa Cruz CA	SC-398	1:500 *	rabbit
TGFβrII	Santa Cruz	SC-400	1:100 *	mouse

^*for immunohistochemistry: diluted in 1.0% bovine serum albumin in a solution of 0.10 M phosphate buffer and 0.90% NaCl with 0.75% Tween

^#for immunoblotting: diluted in 2.5% non-fat dry milk in a solution of 0.10 M phosphate buffer and 0.90% NaCl with 0.10% Tween

All immunofluorescence was visualized using a Leica (Nussloch, Germany) microscope fitted with a confocal laser and associated software (Bio-Rad, Hercules CA).

Immunoblots

Telencephalic tissue was collected from Foxg1^+/+, Foxg1^+/Cre, and Foxg1^Cre/Cre littermates on G13.5. Samples were homogenized by immersion in a lysis buffer and sonication. The lysis buffer was comprised of 1.0% Nonidet P-40, 0.50% deoxycholic acid, 0.010% sodium dodecylsulfate (SDS), protease inhibitor cocktail (Sigma), 1.0 mM sodium orthovanadate, and 10 mM sodium fluoride in 0.010 M phosphate buffered saline. After centrifugation at 10,000 rpm for 10 min, the protein content of the supernatant was determined using a BioRad protein assay.

Supernatants (60 μg/sample) and molecular weight markers (Amersham, Piscataway NJ), were loaded on 12% SDS-polyacrylamide gels, separated by electrophoresis, and transferred to nitrocellulose membranes. Non-specific immunoreactivity in the membranes was blocked with a wash in 5.0 % non-fat dehydrated milk (NFDM) in 0.10% Tween and PBS. Separated proteins were probed for total Smad2/3 and pSmad2 by incubation overnight at 4°C with a primary antibody (Table 1). After washing, the blots were incubated with a horseradish peroxidase-conjugated secondary antibody directed against mouse IgG (1:1000 in 2.5% NFDM; Amersham) or rabbit IgG (1:3000 in 2.5% NFDM; Amersham) for 45 min. Immunotagged protein bands were visualized using a chemiluminescent detection reagent (Amersham). Membranes were then stripped of immunolabel and re-probed for actin expression. The amount of actin expression was used as a loading control.

Quantitative analyses

Quantitative analyses of anatomical and biochemical samples were based on multiple preparations obtained from multiple litters. Means of data were calculated for slices taken from animals in a single litter. Grand means and the associated variations among litters of particular genotype or treatment group were used for statistical analyses.

The numbers of p21- and p73-immunopositive cells were determined in slices obtained from wild-type embryos on G13.5 in each of the treatment groups described above. Confocal images of these preparations (triple-immunolabled for p21, p73, and Ki-67) were captured at 40x magnification; the images included full cross-sections of the cortical hem. The numbers of p21+ and p73+ cells in the hem were quantified from these captured images. At least three slices from each of four separate litters (thus, n=4) were analyzed for each treatment condition.

Densitometric analysis of the immunoblots was performed using an Image Station (Kodak, Rochester NY). Variations in total protein loaded on the gels were normalized using the amount of actin expression as a standard. The amount of pSmad2 in each sample was determined in relation to the amount of total Smad2/3. Samples from three separate litters (n=3) were analyzed for each genotype.

Differences among treatment groups were assessed with Tukey tests for multiple comparisons. Sample variations reported in the text and on all graphs are standard errors of the means. The number of samples used in each analysis is described in the Results.

RESULTS

Transient p21 expression by newly generated CR neurons

On G12.5 and G13.5, the expression of reelin and p73, CR neuron-specific markers, was observed to four disparate sites in the telencephalic neuroepithelium: the septum, cortical hem, strionuclear neuroepithelium (SN), and caudomedial wall of the telencephalon (Figs. 1, 2, and 3). The spatiotemporal patterns of p21 expression and co-expression of p21 with CR neuronal markers and of the proliferation marker Ki-67 were examined at these sites.

Figure 1. Focal p21 expression in newly generated CR neurons.

In wild-type fetal forebrain on G12.5, numerous p21+ cells were concentrated in the upper septum and cortical hem, whereas p21 expression was largely absent in adjacent regions (A and B). The dorsal septum contained both p21+/reelin− cells (green arrows) as well as p21+/reelin+ cells (yellow arrows) (A′; magnified upper dashed box in A). In reelin+ populations, the number of p21+/reelin+ cells decreased with distance from their origin site (e.g., the dorsal septum and cortical hem) (A″) and (B′; reelin+/p21+, white arrows; reelin+/p21-, yellow arrows). Scale bars are 200 μm (A and B) and 50 μm (A′, A″, and B′). Dotted lines depict the ventricular surface. LV, lateral ventricle.

Figure 2. Co-localization of p21 and p73 in the SN, septum, and cortical hem of a wild-type mouse on G13.5.

p21 (green) and p73 (red) were mostly co-expressed in newly generated CR neurons in and adjacent to the Ki-67+ (blue) neuroepithelium however p73, but not p21, expression was retained in CR neurons with distance from the origin site (A, B, and C; white arrows). Some p21+ cells in the cell cycle (Ki-67+) also co-expressed p73 (A′, B′, and C′; white arrows), however a few p21+/Ki-67+ only cells were present (blue arrows). Post-mitotic (Ki-67−) CR neurons expressing both p21 and p73 were concentrated next to the neuroepithelium (A′, B′, and C″; yellow arrows). A and A′. SN; B and B′. Septum; C and C′. Cortical hem. Scale bars are 200 μm (A and D) and 100 μm (A′, B, C and D′). Dotted lines depict the ventricular surface (VS). LV, lateral ventricle.

Figure 3. Foxg1 expression is absent in regions of CR neuronal production.

Foxg1 immunolabeling (blue) in wild-type fetal brains on G12.5 was largely absent in regions of strong p21 (green) and reelin (red) immunoexpression in the cortical hem (A) and caudomedial wall (B). In Foxg1^Cre/Cre brains on G12.5, reelin+ cells made up most of the post-mitotic population and p21 immunoreactivity is highest in the dorsal (C′) and ventral (C″) portions of the forebrain (C). Magnified views of the p21+ regions in the forebrain of a Foxg1-null fetus revealed a similar distribution of p21+ cells in the cerebral wall e.g., post-mitotic p21+/reelin+ (white arrows) and p21+/reelin− (green arrows) (C′ and C″). Scale bars are 200 μm (A, B, and C) and 50 μm (C′ and C″).

On G12.5, both reelin-positive (reelin+) and p21-positive (p21+) cells were concentrated at the upper limit of the septal enlargement and within the hem (Fig. 1A and 1B). Closer examination of the septum revealed numerous strongly labeled p21+ cells situated at the top of the septal enlargement whereas fewer, less intensely labeled p21+ cells were observed ventrally (Fig. 1A ′ and 1A ″). In the dorsal aspect of the septum, there was a mixing of p21+/reelin− cells and cells co-expressing p21 and reelin (p21+/reelin+). Most ventral p21+ cells co-expressed reelin. A similar pattern of reelin/p21 immunoreactivity was observed in the cortical hem in that intense p21 immunoreactivity was evident in the hem itself, but nuclear p21 expression diminished in reelin+ cells in the marginal zone with distance from the hem (Fig. 1B ′). The close proximity of p21+/reelin− and p21+/reelin+ cells in the septum, an established birthplace of CR neurons, suggests that p21 expression may precede reelin expression. Further, the gradual decrease in p21 immunolabeling in reelin+ CR neurons within the known CR neuron migratory paths from the septum (ventral; Bielle et al., 2005) and cortical hem (neocortical marginal zone; Takiguchi-Hayashi et al., 2004, Garcia-Moreno et al., 2007), indicates that p21 may be down-regulated by reelin+ cells as they migrate from their respective birthplaces.

Combined immunolabeling for p73 (a nuclear marker of CR neurons expressed before reelin) and the cell proliferation marker Ki-67 was used to examine the timing of p21 expression in the neuroepithelium in three CR neuron generating areas (Fig. 2). In the neuroepithelia of the septum, cortical hem, and SN on G13.5, some p21+ cells co-localized with Ki-67 though most were Ki-67- and thus considered to be post-mitotic (Fig. 2A ′, 2B ′, and 2C ′; Supplementary Fig. 1). This concurs with data from a recent paper (Siegenthaler and Miller, 2005) that showed that p21 expression is initiated during the G1 phase after cells pass through their final mitotic division and persists in post-mitotic neurons within the cortical VZ.

p21 was co-localized with p73 in both cycling (Ki-67+) and post-mitotic (Ki-67minus;) CR neurons in and adjacent to the source neuroepithelium (Fig. 2A ′, 2B ′, and 2C ′; Supplementary Fig. 1). Both the frequency of p21+ cells and the intensity of p21 immunolabeling declined with distance from the generation site, however, the p73+ immunoreactivity was retained (Fig. 2A, 2B, and 2C).

Regions of CR neuronal generation are permissive for TGFβ signaling

TGFβ initiates p21 transcription in numerous cell types and Foxg1, which is expressed throughout the telencephalon, acts as a potent inhibitory signal of TGFβ-dependent transcription of p21. Thus, it is likely that Foxg1 expression or activity is reduced in areas of p21 expression. Indeed, triple immunolabeling for p21, reelin, and Foxg1 showed that Foxg1 protein expression was not apparent in the p21/reelin-enriched cortical hem (Fig. 3A) and caudomedial telencephalic wall (Fig. 3B).

The distribution of p21, reelin, and Foxg1 was examined in Foxg1^Cre/Cre mice. In the Foxg1^Cre/Cre mice, regions of p21 immunoreactivity were expanded (Fig. 3C) as compared to wild-type mice. Even so, concentrations of p21+ cells were highest in the dorsal and ventral limits of the Foxg1-null forebrains and very few p21+ cells were observed in the central region where reelin staining was highest. In these dorsal (Fig. 3C ′) and ventral (Fig. 3C ″) regions, the p21+/reelin+ cells were distributed as observed in wild-type brains in that a few p21+/reelin− cells were observed in the neuroepithelium and p21+/reelin+ cells were concentrated adjacent to the neuroepithelium.

A functional TGFβ signaling pathway requires TGFβrI, TGFβrII, and pSmad2. On G12.5, both receptors were expressed throughout the cerebral wall, including the cortical hem (Figs. 4A and 4B). Whereas TGFβrI expression was evenly distributed in the proliferative and post-mitotic compartments, TGFβrII was more richly expressed by post-mitotic cells. Cortical hem cells were immunopositive for pSmad2, and p21+ cells in the cortical hem co-labeled with pSmad2 (Fig. 4C). Interestingly, in the absence of Foxg1, TGFβrI and TGFβrII immunolabeling in the forebrain appeared more intense and more expansive (Fig. 4D and 4E). Further, the forebrains of Foxg1^Cre/Cre mice contained significantly (p=0.019; n=3 for 3 litters including 9 animals) more pSmad2 than did heterozygous or wild-type littermates (Fig. 4F). Thus, (1) TGFβ signaling pathways are expressed in Foxg1-poor CR generation sites where p21 expression is highest and (2) TGFβ signaling activity is elevated in the forebrains of Foxg1-null mice in which both p21 expression and CR neuron generation is increased. Collectively, this evidence indicates that TGFβ signaling, in the absence of its inhibitor Foxg1, may stimulate CR neuron generation through upregulation of p21.

Figure 4. Intact TGFβ signaling pathway in an area of active CR neuronal generation.

TGFβrI (A and A′) and TGFβrII (B and B′) and their downstream signaling target p-Smad2/3 (C′) were evident in the cortical hem on G12.5. Further, pSmad2 (green; C′) and p21 immunoreactivity (red; C) were co-localized (yellow; C″). In the forebrain of Foxg1^Cre/Cre mice, the intensity of TGFβrI (D) and TGFβrII (E) immunolabeling was qualitatively increased. Quantitative analysis of immunoblots show that though the amount of Smad was similar in wild-type and Foxg1 heterozygotes and knockouts, there was a p-Smad2/3 protein expression was a significant (p<0.05) increase in p-Smad in Foxg1 knockouts as compared to wild-type and heterozygote littermates (F). Scale bars are 200 μm (A, B, D, and E), 100 μm (A′ and B′), and 50 μm (C, C′, and C″). LV: lateral ventricle; VS: ventricular surface.

To directly test whether TGFβ signaling was required for p21 expression and CR neuronal generation in the forebrain, cortical explants from 13.5-day-old mouse fetuses were treated with TGFβ1 (40 ng/ml) or a TGFβrI inhibitor (SB431542; 100 μM) and then tissue was examined and quantified for cells immunoexpressing p21 and p73. This analysis focused on one site of CR neuron generation, the cortical hem.

Control explants appeared similar to in vivo tissue. That is, p21+ cells were in and about the cortical hem and p21 immunolabeling diminished in p73+ cells with distance from the tip of the cortical hem (Fig. 5A; Supplementary Fig. 2). Treatment with TGFβ1 significantly (p<0.001; n= 30 slices from 4 litters) increased the number of p21+ cells, p73+ cells, and co-expressing cells (Figs. 5B and 5D; Supplementary Fig. 2). On the other hand, exposure to SB431542 significantly (p<0.001; n= 28 slices from 4 litters) decreased the numbers of both p21+ and p73+ cells (Figs. 5C and 5D; Supplementary Fig. 2).

Figure 5. Altered TGFβ signaling affects CR neuronal generation in the cortical hem.

The cortical hem cultured from a 13.5-day-old wild-type fetus appeared similar to in vivo cortical hem in that most p21+ cells (green) and p73+ cells (red) did not express Ki-67 (blue) (A). Treatment with TGFβ1 (40 ng/ml) increased the number of p21+/p73+ cells (B), whereas treatment with SB431542 (100 μM) decreased the p73+/p21+ population in the explant cortical hem (C). The graph (D) represents the number of p21+ (left) and p73+ (right) cells in the hem of control, TGFβ1- or SB431542-treated cortical hem explants. The single asterisks and cross denote a statistically significant increase and decrease relative to control, respectively. Scale bars are 50 μm.

Nuclear shuttling of Foxo3a and p21 expression in CR neurons

Immunolabeling studies were performed to determine whether the transient expression of p21 correlated with Foxo3a nuclear shuttling in young CR neurons. Sections at the level of the cortical hem and SN from 13.5-day-old fetuses were triple immunolabeled with anti-p21, Foxo3a, and reelin antibodies in most telencephalic structures. In the majority of the forebrain tissue, Foxo3a was cytoplasmic (Fig. 6A). Closer examination of both the cortical hem and SN revealed that p21+ cells (both reelin− and reelin+) exhibited nuclear Foxo3a (Fig. 6B). In contrast, reelin+/p21− cells distant from the generation site expressed cytoplasmic Foxo3a (Fig. 6C). Thus, nuclear localization of Foxo3a paralleled expression of p21 in newly generated CR neurons.

Figure 6. Shuttling of Foxo3a coincides with p21 upregulation in CR neuronogenetic sites.

A. Foxo3 immunoreactivity (blue) was cytoplasmic in most telencephalic regions on G13.5. Reelin+ cells (red) were concentrated in the cortical hem and SN. p21+/reelin− cells and p21+/reelin+ cells immediately adjacent to the source neuroepithelium contained nuclear Foxo3a (A″. p21+/reelin−/nuclear Foxo3a+, yellow arrows; p21+/reelin+/nuclear Foxo3a+, white arrows), whereas p21−/reelin+ cells ventral to this site contained cytoplasmic Foxo3a (blue arrows). B. Higher magnification of the cortical hem in A′ (lower dashed line box), showed numerous nuclear Foxo3+ cells in the cortical hem (white and yellow arrows) and that all reelin+ cells contained nuclear Foxo3a (white arrows). The same image with p21+ immunoreactivity revealed that all p21+ cells contained nuclear Foxo3a (B′). C. Distant from the cortical hem (upper dashed line box in A′), reelin and Foxo3a immunoreactivity co-localized in the cytoplasm (C; blue arrows) and these cells were not p21+ (C′). A few reelin+ cells had nuclear Foxo3a (white arrows) and were p21+ (C′). D–E. Foxo3a immunoreactivity in Foxg1^+/+ and Foxg1^Cre/Cre lateral neocortex on G17.5 showed that Foxo3a was entirely cytoplasmic, even in cells that were p21+ (D′ and E′). Scale bars are 200 μm (A), 100 μm (A′ and A″), 10 μm (B–C)

Although it appeared that Foxo3a was involved in p21 expression during the birth of CR neurons, it did not establish whether nuclear Foxo3a always coincided with, and thus was potentially required for, p21 expression. To examine this, brains from 17.5-day-old wild-type fetuses were triple-immunolabeled for p21, Foxo3a, and proliferating cell nuclear antigen (PCNA; a marker of cycling cells). Unlike p21+ cells in the cortical hem and SN, Foxo3a was in the cytoplasm of p21+ neural progenitors in the VZ (Fig. 6D). This was also the case in the neuroepithelia of Foxg1^Cre/Cre mice on G17.5 wherein p21+ cells were more prevalent, yet none appeared to contain nuclear Foxo3a (Fig. 6E). Thus, co-incident expression of nuclear Foxo3a and p21 in neurons apparently was limited to the generation of CR neurons.

To ascertain whether the IGF-1/PI3-K pathway was responsible for Foxo3a nuclear translocation in CR neurons, explants containing the cortical hem were treated with IGF-1 (300 ng/ml), LY-294002 (60 μM; an inhibitor of PI3-K activity), or SB431542 (100 μM). Treatment with IGF-1 or LY-294002 did not affect Foxo3a nuclear localization or the number of p21+ cells (Figs. 7B and 7C; Supplemental Fig. 3). Thus, Foxo3a nuclear shuttling was not controlled by the IGF-1/PI3-K pathway in CR neurons. On the other hand, explants of the cortical hem treated with the Smad inhibitor SB431542 not only had fewer p21+ (and reelin+) cells than the other treatment conditions, but many hem cells still contained strong Foxo3a nuclear localization and no p21 immunoreactivity (Fig. 7D; Supplemental Fig. 3), suggesting that Foxo3a nuclear shuttling occurs independent of TGFβ signaling. Therefore, Foxo3a and TGFβ/Smad signaling pathways likely work in parallel to drive transcription p21 expression during the generation of CR neurons.

Figure 7. Foxo3a localization is unaltered following application of modulators of the IGF/Pi3K and TGFβ pathways.

Treatment of cortical explants from 13.5-day-old fetuses with IGF-1 (300 ng/ml) (G) or the PI3-K inhibitor LY-294002 (60 μM) (H) did not affect p21 expression or the location of Foxo3a immunoreactivity in the cortical hem. Explants treated with SB431542 (I) had diminished p21 expression, but contained numerous cells expressing nuclear Foxo3a without p21 expression (white arrows). Scale bars are 50 μm.

DISCUSSION

CR neurons are an early-born, specialized type of neuron that is derived from specific regions in the telencephalic neuroepithelium. Previous studies established that Foxg1, a potent inhibitor of both CR neuronal fate and TGFβ signaling, is important for confining the birth of CR neurons to discrete sites. The present study shows (1) that p21 expression is coincident with the birth of CR neurons in Foxg1-weak regions of the forebrain and (2) that TGFβ signaling stimulates the generation of CR neurons in the cortical hem and this correlates with up-regulation of p21. Furthermore, the present study identifies a potential novel role for a second member of the Fox family, Foxo3a, in CR neuronal generation. Specifically, nuclear localization of Foxo3a coincides with the up-regulation and loss of p21 expression in emerging CR neurons.

Function of transient p21 expression in CR neuronal production

Previous investigations of p21 transcript expression in the developing forebrain identified p21+ cells in one site of active CR neuronal generation, the cortical hem (Mallamaci et al., 2000; Seoane et al., 2004). These data concur with a profiling study by Yamazaki and colleagues (2004). They showed that p21 was among the transcripts that were highly expressed by immature CR neurons in the cortical hem on G13.5. In contrast, “mature” CR neurons in the marginal zone do not express p21 (Yamazaki et al., 2004). Not only is our detection of p21+ cells in sites of CR neuronal generation consistent with these studies, it provides additional insight into the function of p21 in CR neuronogenesis.

p21 is a potent inhibitor of cell cycle progression of cells in the cortical ventricular zone (VZ) (Siegenthaler and Miller, 2005). That is, p21 is mostly expressed by post-mitotic neurons. Based on this, it was posited that p21 induction promotes the birth of new CR neurons. Indeed, the paucity of p21+/Ki-67+ cells in the neuroepithelium of the septum, cortical hem, and SN suggests that cell cycle exit is rapid once p21 expression is initiated. The effectiveness with which p21 forces cells out of the cell cycle makes it an attractive initiator of CR neuronal generation, particularly because the entire telencephalic complement of CR neurons must be generated in a relatively short time period (less than three days).

The short duration of p21 expression in CR neurons suggests (1) that it is unnecessary for the function of mature CR neurons, (2) that it is involved in the definition of CR neuronal fates, and (3) that the pathway that drives p21 expression cannot be maintained, indeed, need not be maintained once the progenitors for CR neurons have irrevocably exited the cell cycle. Though nuclear localization of Foxo3a may be important for driving p21 expression in CR neurons, translocation of Foxo3a into the nuclei of adult hippocampal neurons activates of apoptotic pathways (Brunet et al., 1999; Zhang et al., 2002). Consequently, long-term nuclear expression of Foxo3a may be harmful to CR neurons. Thus, the rapid shuttling of Foxo3a out the nucleus after CR neuronal definition may be important for preventing activation of cell death pathways.

p21 is not involved in the generation of all CR neurons. No focal p21 expression is evident at the pallial-subpallial boundary, a source of CR neurons (Bielle et al., 2005). This region, nestled among neocortical and striatal precursors, differs from other sites of CR neuronal generation because it is enriched with Foxg1. Foxg1 deficiency is permissive for TGFβ-regulated cycling activity. Inasmuch as TGFβ1 promotes telencephalic cells to exit via p21 induction (Siegenthaler and Miller, 2005), it is unclear what signal initiates the exit of CR neurons from the pallial-subpallial boundary. Conceivably, these CR progenitors may require a more cell-autonomous cell cycle exit cue. This is a conundrum because many neighboring neuronal progenitors remain in the cell cycle.

It is important to recognize that p21 alone is not instructive for CR neuronal identity. p21 works in concert with other site-specific cues to produce CR neurons. This conclusion is supported by evidence that p21+ cells in the VZ of older wild-type fetuses, e.g., on G17.5, are not fated to become CR neurons. Even in the Foxg1-null mice, where p21 expression is abundant at all fetal stages, CR neuronal generation is limited to early forebrain development (Muzio and Mallamaci, 2005). Thus, p21 is not an exclusively fate cue for CR neurons, but rather an efficient means to exit the cell cycle.

TGFβ signaling in CR neurons

There is precedent for TGFβ regulating cell differentiation via p21 in non-neural cell types (Seoane et al., 2004), however, only a few examples exist in the developing CNS. Secreted TGFβ2 in primary cerebellar culture medium, as well as brain-derived neurotrophic factor, are responsible for increases in the expression of the CKIs p21 and p27 (Lu et al., 2005). In turn, these CKIs promote the differentiation of cerebellar neurons. In cortical explants, treatment with exogenous TGFβ1 increases p21-dependent cell cycle exit of VZ progenitors (Siegenthaler and Miller, 2005).

A unique feature of most sites of CR neuronal generation is the absence of Foxg1 expression. Presumably, this expression pattern permits, amongst other things, TGFβ signaling to proceed unhindered in these areas, thus, limiting TGFβ-directed CR neuronal generation to the regions. Even in the Foxg1^Cre/Cre mice, however, p21 expression and CR neuronal generation are limited to the most dorsal and ventral neuroepithelia. Thus, something beyond the absence of Foxg1 and intact TGFβ signaling is required for the p21-dependent generation of CR neurons. The implication is that restricting elements beyond Foxg1 control CR neuronal differentiation. Ideal candidates are signaling elements that direct nuclear localization of Foxo3a.

Fox family proteins in the generation of CR neurons

It has been hypothesized that Fox transcription factors play a role in lineage definition in developing tissues (Lehmann et al., 2003). CR neurons are an example wherein two Fox family proteins may work in opposition in that Foxg1 is known to inhibit TGFβ-dependent transcription of p21 whereas nuclear Foxo3a promotes p21 expression (Fig. 8). The original description of this TGFβ/Fox pathway (Seoane et al., 2004) links the resistance of glioblastoma cells to the anti-proliferative effects of TGFβ1 signaling to the opposing functions of Foxg1 and 3a, however, our work is the first to describe this pathway at work in vivo.

Figure 8. Schematic of TGFβ-Fox signaling pathways in the cortical hem.

A. The generation of CR neurons occurs among neural progenitors (cells with pink perikarya and blue nuclei) in Foxg1-deficient zones. This generation is precipitated by the TGFβ-dependent transcription of p21 (green circles in the nuclei) and the nuclear localization of Foxo3a. B. After the cells exit the proliferative population, the young CR neurons (cells with red perikarya and yellow nuclei) upregulate their p21 expression. C. As these neurons (cells with red perikarya and orange nuclei) migrate from the generation site (e.g., the cortical hem), Foxo3a translocates to the perikaryon and TGFβ signaling can no longer maintain p21 expression). D. In Foxg1-rich portions of the proliferative zones where CR neurons are not readily generated (i.e., below medial and lateral cortex), Foxg1 expression actively suppresses TGFβ-induced transcription of p21 and Foxo3a is cytoplasmic in neuroepithelial precursors.

Foxo3a, like other Foxo proteins, is a downstream target of the IGF/PI3-K pathway that controls the nuclear shuttling, and thus, transcriptional activity for Foxo proteins (Biggs et al., 1999; Brownawell et al., 2001; Van Der Heide et al., 2004). In light of this evidence, it was surprising that application of exogenous IGF-1 or a PI3-K inhibitor affected neither Foxo3a nuclear localization nor p21 expression in the cortical hem. There may be a novel pathway that regulates Foxo3a nuclear shuttling in CR neurons. A more plausible explanation, however, is that Akt, which phosphorylates Foxo3a directly, is regulated by redundant pathways that include IGF/PI3-K. Interestingly, p21 expression in the neuroepithelium of non-CR producing areas does not correlate with nuclear Foxo3a. Other Foxo proteins, specifically Foxo1 or Foxo4, also bind to Smad proteins and aid in p21 transcription (Seoane et al., 2004). These Foxo proteins may be involved in control of p21 transcription in other portions of the cortical neuroepithelium.

The source of CR neurons in the telencephalon and the role that Foxg1 plays in CR neuronal definition is a topic of much research and debate. The debate centers on whether CR neurons are derived from discrete regions of the neuroepithelium and migrate to the marginal zone or arise as early daughter cells of neocortical VZ progenitors. One key piece of evidence supporting a neocortical origin of CR neurons is the phenotype of the Foxg1-null mice in which CR neuronal production is grossly expanded at the expense of the normal cortical layers. This indicates that Foxg1 (a) suppresses early fate decisions of CR neurons in neocortical VZ progenitors following CR neuronal production and (b) enables the production of excitatory and inhibitory projection neurons (Hanashima et al., 2004). Recent studies, however, attribute the increase in CR neurons in the Foxg1-null forebrain to a gross expansion of medial, CR neuron-producing areas at the expense of lateral and ventral telencephalon (Martynoga et al., 2005, Muzio and Mallamaci, 2005). In light of this and other evidence, Foxg1 emerges as an important patterning protein. Its expression enables early expansion of lateral and ventral neuroepithelium with little or no requirement for neuronogenesis, specifically the production of CR neurons. Thus, generation of the earliest neurons, the CR neurons, is largely confined to Foxg1 non-expressing regions where TGFβ signaling and p21 induction is unhindered and where rapid cell cycle exit and neuronal birth can occur without disruption of progenitor pool expansion.

Supplementary Material

01

Supplementary Figure 1.

The individual channels for Ki-67 (left), p21 (center), and p73 (right) immunolabeling for the merged images provided in Figure 2 have been separated. Yellow arrows indicate cells that are p21+/p73+/Ki-67−, blue arrows denote p21+/p73−/Ki-67+ cells, and white arrows indicated cells expressing all three markers.

02

Supplementary Figure 2.

The images depict the expression of Ki-67 (left), p21 (center), and p73 (right) in explant cultures of cortical slices. Slices were subjected to three treatments: no treatment (top), treatment with TGFβ1 (middle) or SB431542 (bottom). Capped arrows identify cells that express p21, p73, and Ki-67. Arrows denote p21+/p73+/Ki-67− cells and arrowheads identify cells expressing only p73.

03

Supplementary Figure 3.

The cortical hem in slice preparations was treated with control medium, or a medium supplemented with insulin-like growth factor (IGF) 1, the pharmacological blocker LY294002, or SB431542. Tissue was immunostained for Foxo3a (left), p21 (center), or reelin (right) expression. Arrows with and without caps identify cells with cytoplasmic and nuclear Foxo3a expression, respectively.

ABBREVIATIONS

CKI

cyclin-dependent kinase inhibitor

CR

Cajal-Retzius

Foxg1^Cre/Cre

Foxg1-null

Foxg1^+/+

wild-type

G

gestational day

IGF

insulin-like growth factor

PBS

0.10 M phosphate buffered saline

PCNA

proliferating cell nuclear antigen

PI3-K

phosphotidylinositol 3-kinase

pSmad

phosphorylated

Smad SN

strionuclear neuroepithelium

TGFβ

transforming growth factor β

TGFβrI

transforming growth factor β receptor type I

TGFβrII

transforming growth factor β receptor type II

VZ

ventricular zone