BMP2 and FGF2 Cooperate to Induce Neural-Crest-Like Fates from Fetal and Adult CNS Stem Cells

Research Article 5849 BMP2 and FGF2 cooperate to induce neural-crest-like fates from fetal and adult CNS stem cells

Martin H. M. Sailer1,2,3, Thomas G. Hazel1, David M. Panchision1,4, Daniel J. Hoeppner1, Martin E. Schwab2,3 and Ronald D. G. McKay1,* 1Laboratory of Molecular Biology, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, MD 20892, USA 2Brain Research Institute, University of Zurich, 8057 Zurich, Switzerland 3Department of Biology, Swiss Federal Institute of Technology, 8057 Zurich, Switzerland 4Center for Neuroscience Research, Children’s Research Institute, Children’s National Medical Center, Washington DC, 20010, USA *Author for correspondence (e-mail: [email protected])

Accepted 21 September 2005 Journal of Cell Science 118, 5849-5860 Published by The Company of Biologists 2005 doi:10.1242/jcs.02708

Summary CNS stem cells are best characterized by their ability to stem cells from E14.5 cortex, E18.5 cortex and adult self-renew and to generate multiple differentiated subventricular zone, but with a progressive shift toward derivatives, but the effect of mitogenic signals, such as gliogenesis that is characteristic of normal development. fibroblast growth factor 2 (FGF2), on the positional identity These data indicate that FGF2 confers competence for of these cells is not well understood. Here, we report that dorsalization independently of its mitogenic action. This bone morphogenetic protein 2 (BMP2) induces rapid and efficient induction of dorsal fates may allow telencephalic CNS stem cells to fates characteristic of identification of positional identity effectors that are co- neural crest and choroid plexus mesenchyme, a cell type of regulated by FGF2 and BMP2. undetermined lineage in rodents. This induction occurs both in dissociated cell culture and cortical explants of embryonic day 14.5 (E14.5) embryos, but only when cells have been exposed to FGF2. Neither EGF nor IGF1 can Supplementary material available online at substitute for FGF2. An early step in this response is http://jcs.biologists.org/cgi/content/full/118/24/5849/DC1 activation of ␤-catenin, a mediator of Wnt activity. The CNS stem cells first undergo an epithelial-to-mesenchymal Key words: Forebrain, Neural stem cell, Cranial neural crest, transition and subsequently differentiate to smooth-muscle Choroid plexus mesenchyme (CPm), Epithelial-mesenchymal and non-CNS glia cells. Similar responses are seen with transition (EMT), Snai1, Snai2 Journal of Cell Science

Introduction midline of the neural tube (Lee and Jessell, 1999b; Meulemans Clonal analysis shows that stem cells can be derived from the and Bronner-Fraser, 2004). In the anterior neural tube, central nervous system (CNS) and maintained in culture by the including the telencephalon, choroid plexus (CP) is the dorsal- mitogenic actions of FGF2 (Johe et al., 1996). FGF2 and most region. BMP signaling is both necessary and sufficient epidermal growth factor (EGF) are the only known growth for the generation of the CP (Hebert et al., 2002; Panchision et factors that, by themselves, can drive the mitogenic expansion al., 2001). In the more posterior neural tube, BMPs induce roof of neural precursor cells in vitro (Panchision and McKay, plate and neural crest cells (Lee and Jessell, 1999a). Neural 2002). The use of FGF2 as an exogenous mitogen is supported crest cells are a specialized dorsal cell type, unique to by its actions in vivo. FGF2 is expressed together with FGF1 vertebrates, that delaminate from the neural and/or non-neural in the ventricular zone of the developing cortex (Dono, 2003; ectoderm border from regions posterior to the mid- Grove and Fukuchi-Shimogori, 2003). Mice lacking FGF2 diencephalon, and then migrate to distant sites in the embryo show a cerebral cortex size that is diminished by about 45%, (Bronner-Fraser, 2002; Knecht and Bronner-Fraser, 2002; affecting both neurons and glia cells (Vaccarino et al., 1999). Trainor et al., 2002; Wu et al., 2003). The neural crest gives However, it has also been proposed that FGFs can act as rise to peripheral nervous system (PNS) derivatives such as ventralizing signals in mouse cortical explants and cultured peripheral neurons and Schwann cells, along with non-neural neural precursors (Gabay et al., 2003; Kessaris et al., 2004; derivatives such as melanocytes, craniofacial chondrocytes, Kuschel et al., 2003). osteocytes and perivascular cells (smooth muscle, pericytes, Cell fate in the developing CNS is speciﬁed in part by connective tissue) (Gammill and Bronner-Fraser, 2003; Le growth factors that are localized in distinct dorsal and ventral Douarin and Kalcheim, 1999; Meulemans and Bronner-Fraser, organizer domains (Lee and Jessell, 1999b; Wilson and 2004; Trainor et al., 2002; Wu et al., 2003). Rubenstein, 2000). Bone morphogenetic proteins (BMPs) are Previous work has shown that BMPs promote the generation secreted factors expressed in the prospective epidermis at the of mesenchymal derivatives with neural-crest-like phenotypes lateral edges of the neural plate, then subsequently in the dorsal from CNS precursors in vitro (Gajavelli et al., 2004; Mujtaba 5850 Journal of Cell Science 118 (24)

et al., 1998; Rajan et al., 2003). This is in contrast to other recombinant mouse noggin at 2.5 ng/ml; rhIGF1 at 330 ng/ml; bovine studies that have shown CNS neuronal or glial differentiation pancreatic insulin at 25 ␮g/ml (Sigma). To promote neuronal survival after BMP treatment (Gross et al., 1996; Li et al., 1998; Mehler and maturation, we used B27 supplement (1:100 dilution, Invitrogen), et al., 2000), suggesting a context-dependent component to the 40 nM retinoic acid (Sigma), 10 ng/ml nerve growth factor (rhNGF), ␤ BMP response. In this study, we show that exposure to FGF2 10 ng/ml heregulin 1 (rhHRG), 10 ng/ml recombinant rat brain- is required for BMPs to generate neural-crest-like cells in derived growth factor (BDNF) and 10 ng/ml glial-cell-line-derived neurotrophic factor (rrGDNF). cortical explants or cultured stem cells. Co-treatment with FGF2 and BMP2 rapidly upregulates ␤-catenin, a mediator of Wnt activity (Moon et al., 2004) and of Bmp2 itself, suggesting Culture of cortical explants a positive feedback of BMP-signaling. This is consistent with E14.5 rat telencephalon was dissected to be completely free of a role for Wnt-signaling in regulating dorso-ventral identity in meninges. A cortex section of about 800 ␮m by 3200 ␮m along the the developing telencephalon of chick and mouse (Backman et length of the medial ganglionic eminence (MGE) in a distance of al., 2005; Gunhaga et al., 2003). FGF2 and BMP2 treatment about 800 ␮m from the MGE and starting at the anterior MGE pole initially induces genes associated with the epithelial- was dissected (Fig. 1). This section was used for explants of 400-800 ␮ mesenchymal transition (EMT) to a neural-crest-like state m in diameter. The explants were generated by cutting the cortex (Nieto, 2002), followed by terminal differentiation into tissue section into smaller pieces with a microsurgical needle (tungsten); they were grown in the same conditions as CNS stem cells derivatives such as smooth muscle and non-CNS glia. (see below), except for the omission of insulin, progesterone, Forebrain stem cells from multiple ages retain the competence putrescine and selenium from the medium (basal medium). for this dorsal respeciﬁcation. Peripheral neurons are not Transferrin, an iron-binding protein necessary for efficient iron generated under these conditions, but can be induced at low metabolism, was maintained in culture because its removal caused frequency when stem cells are co-treated with the growth retardation and some cell death (data not shown). Medium posteriorizing factor retinoic acid. Thus, although these cell with growth factors was replaced every other day. The explants were types are characteristic of neural crest, the anterior origin of washed twice with basal medium between the ﬁrst and second phase the responding cells biases the FGF-BMP-Wnt-induced dorsal of treatment (see Table 1). The responses to growth factors were transition to a phenotype most similar to cranial mesenchyme recorded by phase-contrast microscopy using a Zeiss Axiovert 10 or choroid plexus mesenchyme (CPm). microscope (Carl Zeiss Inc.).

Culture of CNS stem cells Materials and Methods Both fetal and adult CNS stem cells were isolated and cultured as Growth factors previously described (Johe et al., 1996; Kim et al., 2003) in Dulbecco’s Growth factors were used at the following concentrations (all from modifies Eagle’s medium (DMEM) F12 with N2 supplements (Kim et R&D Systems with BSA as carrier protein, if not otherwise stated): al., 2003), unless otherwise noted. We used E14.5- (Taconic E15) and recombinant human (rh) FGF2 at 10 ng/ml (146 aa) for fetal and at E18.5- (Taconic E19) timed pregnant Sprague Dawley rats (Taconic, 20 ng/ml for adult stem cells; rhBMP2, BMP4, BMP7 at 20 ng/ml; Germantown, NY; noon of day of plug detection equals E0.5). Adult rats Journal of Cell Science Fig. 1. Expression of p75NGFR, SMA and GFAP in the E14.5 rat forebrain and cranial mesenchyme. Fluorescent antibody staining of E14.5 rat embryo at the telencephalic level. Illustration at upper left orients panels A-G. Red rectangle with asterisk illustrates the cortical region dissected for all experiments. (A) p75NGFR (green) and smooth muscle ␣-actin (red) show little or no expression in brain but high levels of p75NGFR in cranial mesenchyme and peripheral ganglia. Arrow indicates choroid plexus. (B-D) Higher magnification, showing (B) cranial mesenchyme, (C) choroid plexus (arrow) and confocal image of choroid plexus epithelium (CPe) and (D-F) mesenchyme. Notice that some CPm cells are p75NGFR+/SMA+, whereas the CPe shows fainter staining for these markers. (G) CPe shows moderate co-expression of p75NGFR (green) and GFAP (red), whereas the CPm is strongly p75NGFR+/GFAP+. Cranial mesenchyme is stongly p75NGFR+ but only weakly GFAP+. DAPI (blue) identifies all cell nuclei. Bars, 160 ␮m (A); 40 ␮m (B-C, E); 15 ␮m (D). BMP2 and FGF2 in CNS stem cells 5851

Table 1. p75NGFR induction in E14.5 cortical explants mouse, 1:200, gift of Juan Archelos), smooth-muscle ␣-actin (SMA) Explant count (mouse, 1:800, Sigma), smooth-muscle myosin heavy-chain 1+2 Phase I r Phase II (SMMHC, rabbit, 1:800, gift of Robert Adelstein), Sox9 (rabbit, + + Experiment 4 days 5 days Total p75 % p75 1:50, Chemicon), peripherin (mouse, 1:100, Chemicon; rabbit, A–r –35001:2000, Chemicon), tyrosine hydroxylase (mouse, 1:1000, Sigma; BFr F5500rabbit 1:400, PelFreez). For ␤-catenin staining, cells were fixed in CIr I52004% PFA containing 4% sucrose at 4°C for 5 minutes, then in 100% D F + I r F + I 49 0 0 r methanol at –20°C for 15 minutes and then extensively washed in E– B3400PBS. Appropriate secondary antibodies (Cy3, 1:300, Jackson FFr F + B 50 50 100 GIr I + B 42 3 7 Immunoresearch, or Alexa Fluor 488, 1:200, Alexa Fluor 555, H F + I r F + I + B 59 59 100 1:300, Molecular Probes) were applied and incubated for 1 hour at room temperature. Nuclei were labeled with 0.25 ␮g/ml 4Ј,6- IEr E2400 r diamidino-2-phenylindole dihydrochloride (DAPI, Sigma) for 15 JE E + B 16 0 0 minutes. Images were photographed, using fluorescent filters, with K F + B r F + B 11 11 100 a Zeiss Axioplan microscope and Zeiss Axiocam HR camera (Carl LBr B800Zeis Inc.) or a Zeiss confocal microscope (LSM 510). Statistical MBr F372978analysis and histogram illustration of cell numbers were performed NN2r N2 23 0 0 using SigmaPlot 5.0 (SPSS Inc.) and Prism 4.0 (Graphpad Software ON2r N2 + B 36 0 0 Inc.). All images were combined for figures using Photoshop 7.0 for Windows (Adobe). Phase I only 5 days PB 1100Western blotting Protein was harvested as described before (Rajan et al., 2003) using Phase I r Phase II 2 days 7 days protease and phosphatase inhibitors. Standardized protein was loaded Q F + B r F 18 18 100 on 4-12% polyacrylamide gels and blotted on nitrocellulose membranes (both Invitrogen); a positive control (Upstate) was E14.5 rat cortical explants were cultured in two phases using DMEM/F12 included. A non-phosphorylated 12-amino-acid (aa) peptide basal medium plus varying combinations of factors as indicated. Explants corresponding to aa 36-44 in human ␤-catenin was used as the were then fixed and screened for p75NGFR expressing cells; explants with immunogen for the activated ␤-catenin antibody. Blots were probed more than one p75NGFR+ cell were scored as positive. Abbreviations: F, FGF2; with HRP-conjugated secondary antibodies (1:10,000, Jackson I, IGF1; E, EGF; B, BMP2; N2, N2 supplements. Concentrations are listed in Immunoresearch) and the antibodies reacted with a chemiluminescent Materials and Methods. reagent (1:2, Pierce).

(8 to 12 weeks old, Sprague Dawley, Taconic) were deeply anesthetized in methoxyﬂurane (Mallinckrodt, Mundelein, IL) before Reverse-transcription PCR analysis decapitation and dissection of the subventricular zone. For adult For gene expression analysis, cultures were harvested in Trizol cultures, noggin was added during FGF2 expansion because we reagent (Invitrogen) for isolation of RNA and treated with DNaseI for

Journal of Cell Science detected low levels of BMP expression in these cells, consistent with 15 minutes at room temperature. The RNA (1000 ng in fetal, 300 ng previous results (Lim et al., 2000). Cultures were maintained in this in adult cell cultures) was reverse-transcribed into first-strand cDNA manner for 5 days; noggin was withdrawn prior to initiating using oligonucleotide (dT)-primers at 0.025 ␮g/ ␮l in a 40 ␮l reaction experiments. (Invitrogen). A 0.5 ␮l aliquot of the first strand reaction in a total Experiments were performed upon passage 1 or 2 using 10 ng/ml volume of 30 ␮l was used for PCR. The following intron-spanning FGF2 and 20 ng/ml BMP2 unless otherwise noted. Cells were primers were used: Snail1 (rat, accession no. XM_342587, 2 cultured at low density (38 cells/cm ) for clonal analysis, medium nucleotides 423-438 and 730-715, 307 bp product), 1.5 mM MgCl2, density (380 cells/cm2) to generate distinct colonies, or high density 57°C annealing; fetal cells: 39 cycles, adult cells: 40 cycles; Snail2 (2546 cells/cm2), all based on previous studies (Rajan et al., 2003). (rat, accession no. AF497973, nucleotides 291-313 and BC062164, To assay early precursor induction, cells were treated with FGF2 ± nucleotides 931-915, 522 bp product), 1.5 mM MgCl2, 57°C BMP2 for varying durations prior to cell fixation or harvesting. To annealing, fetal cells: 34 cycles, adult cells: 34 cycles; Bmp2 (rat, assay differentiated cell types, one treatment paradigm involved 1 day accession no. Z25868, nucleotides 222-242 and 684-662, 463 bp of FGF2 expansion, then 2 days FGF2 ± BMP, then 5 days mitogen product), 1.5 mM MgCl2, 57°C, 40 cycles; GAPDH (rat, accession withdrawal ± BMP2 to maximize post-mitotic cell maturation. A no. AF106860, nucleotides 687-704 and 988-970, 301 bp product), second BMP2 treatment paradigm involved 8 days of BMP2 1.5 mM MgCl2, 57°C annealing, fetal cells: 28 cycles, adult cells: 31 treatment, in varying doses or transient exposure, during continuous cycles. Products were sequence-verified. FGF2 expansion prior to fixation.

Immunocytochemistry Results Embryo explants or cultured cells were fixed with freshly made, cold Expression of p75NGFR, SMA and GFAP in the E14.5 rat paraformalehyde (4%) and processed as described (Panchision et forebrain and cranial mesenchyme al., 2001). Cells were stained with primary antibodies against the Previous results from our lab (Panchision et al., 2001; Rajan et following proteins: Brn3a (rabbit, 1:2000, gift of Eric Turner), al., 2003; Tsai and McKay, 2000) and others (Gajavelli et al., calponin (mouse, 1:500, Sigma), ␤-catenin (mouse, 1:50, Upstate), galactocerebroside (GalC, mouse, 1:200, Chemicon), glial fibrillary 2004; Mujtaba et al., 1998) indicate that cells with neural crest acidic protein (GFAP) (rabbit, 1:600, Dako), nestin (rabbit serum characteristics can be generated from CNS tissue, particularly 130, 1:50 dilution, R.D.G.M.’s laboratory), p75NGFR (mouse, 1:100, in response to BMP treatment. These cells are distinguished by Calbiochem), p75NGFR (rabbit, 1:200, Chemicon), protein zero (P0, the early expression of the low-affinity NGF receptor (NGFR) 5852 Journal of Cell Science 118 (24) (p75NGFR), a transient marker of multipotent neural crest stem cells and some derivatives (Morrison et al., 1999), and the subsequent expression of markers for either smooth muscle (SMA) or glia (GFAP) (Gajavelli et al., 2004; Mujtaba et al., 1998; Rajan et al., 2003; Tsai and McKay, 2000). Interestingly, this in-vitro response appears with CNS precursor cells that are developmentally older than the ages (E8.5- 9.5) at which neural crest is normally generated (Serbedzija et al., 1992), suggesting that these CNS cells possess a latent capacity to generate non-CNS fates. Since BMP signaling in vivo is both necessary and sufficient for generation of CP, the dorsal-most telencephalic identity (Hebert et al., 2002; Panchision et al., 2001), an alternative possibility is that these cells are exhibiting properties of CP rather than neural crest. To identify forebrain cells expressing these markers, the normal expression pattern of p75NGFR and SMA in the E14.5 rat head was defined. The cranial mesenchyme and CP mesenchyme (CPm) both expressed p75NGFR (Fig. 1A-C). The cranial mesenchyme derives from neural crest cells that migrate from regions caudal to the mid-diencephalon (Le Douarin and Kalcheim, 1999); the CPm is adjacent to the choroid plexus epithelium (CPe) Fig. 2. FGF2 and BMP2 induce CNS stem cells to neural-crest-like precursor state. and is of undetermined origin in rodents. The (A) RT-PCR analysis of Msx1 expression in passage 1 of medium-density-plated CPe also expressed p75NGFR but at lower levels stem cells treated with FGF2 or FGF2 and BMP2 after 3 days. GAPDH expression is than in the CPm (Fig. 1D). As expected, shown in reverse-transcribed (+) and non-transcribed (–) samples as loading p75NGFR expression was prominent in other controls. (B) RT-PCR analysis of Snail1 and Snail2 expression under same conditions over 7 days. GAPDH expression is shown in reverse-transcribed (+) and neural crest structures such as peripheral non-transcribed (–) samples as loading controls. (C) Phase images of FGF2- or ganglia. By contrast, SMA expression was FGF2 and BMP2-treated E14.5 rat cortical stem cell cultures at medium density.

Journal of Cell Science NGFR localized to a subset of p75 -positive Notice the initial extension of reticulated processes and cell flattening during BMP2 NGFR+ (p75 ) cells in the head mesenchyme in treatment. (D-G) Comparison of p75NGFR expression in passage 1 cells plated at vascular-like structures intimately associated clonal density; clones marked after 4 days FGF2 expansion (7.7±1.7 cells/clone) with the CPe (Fig. 1D-F). Confocal microscopy were further expanded ±20 ng/ml BMP2 treatment. By 5 days, p75NGFR expression is confirmed the presence of p75NGFR +/SMA+ absent in (D) FGF2-expanded cells but prevalent in (E) FGF2 plus BMP2-treated cells in the CPm (Fig. 1D-F). cells. Percent clones containing any (F) p75NGFR+ cells and total percentage of (G) p75NGFR+ cells during FGF2 expansion without (ᮀ) or with (᭜) BMP2. Graphs GFAP expression was minimal in the E14.5 ␮ ␮ rat forebrain, consistent with the low frequency show the mean ± s.e.m. (n=3). Bars, 80 m (C); 40 m (D,E). of glial differentiation at this early age. CPe expressed GFAP at low levels, but stronger than that seen in more ventral neural epithelium; there were also Vicario-Abejon et al., 2000). Dissociated cells from E14.5 weakly co-expressing GFAP+/p75NGFR+ cells (Fig. 1G). GFAP rat cortex were expanded in FGF2 for 4 days and then expression was higher in the cranial mesenchyme, including passaged and plated at medium density (for details see cells in the CPm that were GFAP+/p75NGFR+. Coexpression of Materials and Methods) just before BMP treatment. Co- GFAP and p75NGFR is characteristic of non-myelinating treatment with BMP2 and FGF2 induced the expression of Schwann cells (Zorick and Lemke, 1996; Zorick et al., 1996). Msx1 (Fig. 2A), which is expressed in dorsal midline cells Thus, CPm cells display characteristics that are between CPe including CPe, and of Snail1 and Snail2 (formerly known as and neural crest derivatives, such as cranial mesenchyme. Snail1 and Slug, respectively) (Fig. 2B), both of which identify CPm and neural crest precursors and are required for proper epithelial to mesenchymal transition (EMT) to a BMP2 and FGF2 efficiently induce CNS stem cells to a migratory phenotype (Aybar et al., 2003; Etchevers et al., NGFR+ + P75 /SNAIL neural-crest-like precursor state 2002; Marin and Nieto, 2004; Nieto, 2002). This co- We next measured the timing and efficiency of CNS stem treatment caused pronounced morphological changes in the cell responses to BMPs. These cells normally give rise to cells, including the formation of reticulated processes and neurons, astrocytes and oligodendrocytes that are eventual cell flattening by 6 days (Fig. 2C). FGF2-BMP2 co- characteristic of their tissue of origin (Johe et al., 1996; treatment also induced p75NGFR, a surface receptor expressed BMP2 and FGF2 in CNS stem cells 5853

Fig. 3. FGF2 is required for BMP2- mediated EMT. Analysis of p75NGFR expression in E14.5 rat cortical explants cultured in basal medium (DMEM-F12 without N2) alone or supplemented with growth factors (see Table 1). Treatments include a 4-day first phase and a 5-day second phase. Low-magnification images show p75NGFR cytoplasmic staining (green) with DAPI+ nuclei (blue); original explants are shown on the right edge of each image. Only media containing FGF2+BMP2 (I- L, P) were able to induce neural crest precursors as measured by p75NGFR expression and migration from explant. Neither insulin nor IGF1 was required (I-K) and IGF was not sufficient (B,D,G) to induce p75NGFR+ precursors. (M,N) Substitution of EGF for FGF2 was not sufficient for neural crest induction. (O,P) A small number of p75NGFR+ cells with thin processes in three of 42 IGF1/IGF1+BMP2 treated explants (O) that differ from the flattened morphologies of FGF2+BMP2/ FGF2+BMP2 treated explants (P). (Q) RT-PCR panel showing Snail2 induction only in explants cultured with both FGF2+BMP2. Abbreviations: F, FGF2; I, IGF1; E, EGF; B, BMP2. Factor concentrations listed in Materials and Methods. Bars, 80 ␮m (A-M); 80 ␮M (N-P). Journal of Cell Science

in CPm and neural crest precursors (Etchevers et al., 2002). FGF2 is required for BMP2-mediated epithelial-to- By clonal analysis, we found that BMP2 co-treatment mesenchymal transition (EMT) generated p75NGFR+ cells in every clone within 4 days (Fig. To determine whether the response of cultured stem cells to 2D-F), indicating that all initially plated and FGF2-expanded BMP2 reflects an EMT, we dissected the same cortical tissue fetal stem cells were responsive to BMP2. The expression of as for our dissociated cultures but instead cultured the whole p75NGFR commenced within 2 days of BMP2 treatment and explants of 400-800 ␮m in diameter. Similar delamination and reached a maximum frequency of about 60% of all cells after migration paradigms using early-gestation neural tube explants 5 days BMP2 treatment (Fig. 2G). By this time, the p75NGFR- have been exploited to prepare neural crest cultures (Kleber et negative (p75NGFR–) cells also had a similarly flattened al., 2005; Lee et al., 2004). The explants were cultured in a morphology, indicating that they also were BMP2- ‘basal medium’ (DMEM-F12 plus transferrin) plus different responsive (data not shown). combinations of additional factors (Table 1). The explants were BMP4 treatment has been shown to induce cell death in first cultured under conditions to manipulate CNS-cell- explants of E10.5 mouse cortical tissue (Furuta et al., 1997), expansion (Phase I, 4 days), and then under conditions to test raising the possibility that the pronounced inductive response EMT (phase II, 5 days), after which the explants were fixed we saw does not reflect the potential of most cortical stem cells. and stained for p75NGFR. No detectable cell death in cortical stem cell cultures was The basal medium by itself did not support cell migration or observed during the first four days of BMP4 treatment, as the induction of p75NGFR (Table 1, experiment A). We then measured by pyknotic cell counting and fragmented nuclei or tested the effect of IGF1 instead of insulin, because IGF1 has staining for activated caspase-3 (data not shown). The fact that a tenfold stronger survival-promoting activity than insulin in BMP2 treatment induced Snail1 and Snail2 expression and that CNS stem cells at the same concentration (Drago et al., 1991). all clones contained p75NGFR+ cells after 4 days of treatment is FGF2-IGF1 co-treatment caused an elaboration of cells from consistent with an efficient dorsal induction of CNS stem cells the explant with a characteristic CNS-precursor morphology to a mesenchymal cell type such as neural crest or CPm (Marin (Fig. 3D) and did not induce p75NGFR expression (Table 1, and Nieto, 2004). experiments B-D). The high density of DAPI+ cells suggested 5854 Journal of Cell Science 118 (24) that this was an overgrowth of CNS precursors. FGF2 or BMP2 cells in three out of 42 cortical explants (Fig. 3G,O; Table 1, alone were not sufficient to induce p75NGFR or migration (Table experiment G); these cells were small and elongated compared 1, experiments E, L, O). Culture of explants with either FGF2 to the large flat p75NGFR+ cells of the FGF2-BMP2-treatments or FGF2-IGF1 in phase I, followed by BMP2 exposure in (Fig. 3P, Table 1, experiments F and H). Additionally, no phase II, strongly induced p75NGFR expression (Fig. 3I-L, Table Snail1/2 expression was seen in IGF1-BMP2 cultures (Fig. 1 experiments F and H) and Snail2 expression (Fig. 3Q). A 3Q), excluding the possibility that they are CPm or neural crest shorter 2-day exposure of cortical explants also induced precursors. Cells expressing p75NGFR were not seen in explants p75NGFR (Fig. 3J, Table 1, experiment Q). The p75NGFR+ cells exposed to EGF, the other known mitogen for cultured CNS migrated away from the explant and adhered tightly to the stem cells (Table 1, experiments I and J). In contrast to surface of the well; this migratory effect was not seen in previously reported experiments carried out in chick explants cultures that did not generate p75NGFR+ cells. Furthermore, at an earlier developmental time point (Garcia-Castro et al., SMA+ cells were found exclusively in these migratory regions 2002), the combination of N2 and BMP2 was not sufficient to (not shown). induce an EMT in rat cortical explants (Table 1N,O). By contrast, treatment with IGF1 alone, followed by IGF1 Furthermore, only the combination of FGF2 and BMP2 plus BMP2 was only able to induce a total of seven p75NGFR+ induced the expression of Snail2 (Fig. 3Q). These experiments indicate that the functions of both BMP2 and FGF2 are required to induce a migratory cell population that expresses genes characteristic of CPm or neural crest.

BMP2-mediated activation of a Wnt signal and BMP2 expression Secreted proteins of the Wnt family have been implicated in the dorsalization of the telencephalon and in the induction and maintenance of neural crest (Backman et al., 2005; Garcia- Castro et al., 2002; Gunhaga et al., 2003; Kleber et al., 2005; Saint-Jeannet et al., 1997). Activation of the canonical Wnt pathway leads to translocation of ␤-catenin to the nucleus and to subsequent transcriptional activation in conjunction with members of the T-cell factor/lymphocyte enhancer binding factor (TCL/LEF) family (Moon et al., 2004). Nuclear ␤- catenin localization was used as a measure for activated canonical Wnt-signaling in FGF2-expanded and BMP2-treated cortical stem cells. FGF2-expanded stem cells express low levels of nuclear ␤-catenin (Fig. 4A,B). BMP2 exposure Journal of Cell Science increased levels of nuclear ␤-catenin signal within 24 hours (Fig. 4C-E). This indicates that BMP-stimulation rapidly activates the canonical Wnt-signaling pathway in NSCs. Since BMP signaling is tightly regulated during neural development, Bmp2 mRNA levels were also measured. Bmp2 mRNA was not detectable in control cultures, but was upregulated within 24 hours of BMP exposure (Fig. 4F). During continuous BMP exposure the Bmp2 mRNA remained elevated (Fig. 4). Furthermore, transient exposure to BMP2 for 2 days was sufficient to induce p75NGFR 7 days later (Table 1Q). Thus, initial BMP exposure might activate a positive feedback loop of BMP signaling.

Fig. 4. BMP2-mediated activation of a Wnt signal and Bmp2 expression. E14.5 rat cortical stem cells were plated at 5092 FGF2 and BMP2 differentiate CNS stem cells to cells/cm2 and exposed to 20 ng/ml BMP2 for 0, 24 or 48 hours. multiple non-CNS derivatives (A) Activated ␤-catenin levels increased after 24 and 48 hours, as Even in the continued presence of FGF2, BMP2-treated cells shown by western blotting. Activated ␤-catenin control-cell lysate eventually stopped proliferating and most adopted a flat and ␣-tubulin were included as references. (B-E) morphology. In low-density cultures, BMP treatment caused ␤ Immunocytochemistry, showing increased -catenin activation after efficient differentiation to cells expressing smooth-muscle ␣- 24 hours of BMP2 treatment; control cells show only few faintly actin (Fig. 5), which is consistent with our previous results positive cells; (B,D) DAPI staining indicates total cell nuclei. (Rajan et al., 2003). This response was specific for the BMP (F) RT-PCR time course analysis of medium-density cultures after ␤ ␤ exposure to FGF2 or FGF2 and BMP2. One day of BMP2 exposure subclass of transforming growth factor (TGF ) factors. was sufficient to upregulate transcription of endogenous Bmp2 Neither FGF2 withdrawal, which is known to induce mRNA. GAPDH expression is shown in reverse-transcribed (+) and differentiation in neural stem cells, nor 8 ng/ml TGF␤1, which non-transcribed (–) samples as loading controls. Bars, 20 ␮m. is known to promote smooth-muscle differentiation in neural BMP2 and FGF2 in CNS stem cells 5855 crest stem cells (Shah et al., 1996), induced any SMA+ cells frequency and maturity of these cells increased in conditions (Fig. 5B-C, M). By contrast, BMP2, BMP4 or BMP7 induced where FGF2 was withdrawn after the initial co-treatment with SMA expression in 92.8%, 96.6% or 90.6% of all cells, BMP2 (data not shown). At doses of 20 ng/ml or higher, the respectively, when treated with concentrations of 20 ng/ml cells assumed a typical sheet-like morphology containing (Fig. 5D-F,M). parallel stress fibers (Fig. 5K-L,N). Even transient exposure to The myogenic potential of cortical stem cells was 20 ng/ml BMP2 was sufficient to direct cells to a smooth- maintained over five passages in culture, corresponding to muscle fate (Fig. 5O). more than 4 weeks in expansion (Fig. 5M). A dose-dependent After treatment with 20 ng/ml BMP2, nearly all cells were effect of BMP2 on differentiation to SMA+ cells was found positive for SMA, calponin and smooth-muscle myosin heavy- when cortical stem cells were co-treated with FGF2 for 8 days. chain (SMMHC) 1 and 2 (Fig. 5P,S-V). These cells also Low doses of BMP2 generated few SMA+ cells, whereas 10 showed nuclear expression of Sox9 (Fig. 5P-R) and p21cip1 ng/ml BMP2 yielded larger numbers of SMA+ cells having a (data not shown) (Panchision et al., 2001), consistent with their small size and immature morphology (Fig. 5G-J,N). Both the identity as post-mitotic smooth-muscle cells. This expression Journal of Cell Science

Fig. 5. BMP-treated CNS stem cells efficiently differentiate into smooth muscle and co-express Sox9, SMMHC1+2 and calponin. (A) Differentiation paradigm 1 (for panels B-F,M) and paradigm 2 (panels G-L,N-V). Cortical stem cells were seeded at low density after first passage. (B-F) Cells differentiated by (B) FGF2-withdrawal or (C) 8 ng/ml TGF␤1 co-treatment are all SMA–. By contrast, 20 ng/ml of BMP2, BMP 4 or BMP 7 efficiently generate SMA+ cells. (M) Quantitation of B-F. Notice that the passage 5 experiment varies from paradigm 1 by using a 7-day co-treatment and 7-day withdrawal. (G-L) Dose-response assay showing no or few SMA+ cells in (G) FGF2-expanded or (H,I) low-dose BMP2 co-treated cells. (J) Percentage of SMA+ cells increases at 10 ng/ml BMP2 but yields immature-looking cells, and plateaus to almost 100% at 20 ng/ml BMP2, even during continued mitogenic expansion (K,L, quantitation in N). (O) Initial transient exposure to 20 ng/ml BMP2 during FGF2 expansion is sufficient to induce SMA+ differentiation by 8 days. (P-V) Cells co-treated with 10 ng/ml FGF2 (F) and 20 ng/ml BMP2 (B) nearly all co-express SMA and Sox9 (Q-R), SMMHC (S-T) and calponin (U-V). Graphs show mean ± s.e.m. (n=3). Bars 40 ␮m (B-L); 20 ␮m (Q-V). 5856 Journal of Cell Science 118 (24) was not seen in untreated cells nor TGF␤1-treated cells (Fig. with a non-myelinating Schwann cell fate (Fig. 6A-C) (Zorick 5Q,S,U and data not shown). and Lemke, 1996). A low proportion (0.7%) co-expressed Apart from smooth-muscle cells, BMP2-treated CNS stem galactocerebroside (GalC) and p75NGFR (Fig. 6D-F), which cells also differentiate to GFAP+ glia (Rajan et al., 2003). In together mark both pro-myelinating and non-myelinating individual colonies, SMA+ cells predominated at the sparse Schwann cells (Jessen and Mirsky, 2002; Zorick et al., 1996). edges, whereas the proportion of GFAP+ cells increased This distinguishes them from oligodendrocytes of the intact dramatically in the dense core (supplementary material, Fig. adult brain that do not express p75NGFR under normal, non- S1), consistent with our previous results (Rajan et al., 2003). injured circumstances (Beattie et al., 2002). Myelin protein Varying the density before treatment with BMP yielded almost zero (MYP0), another gene expressed at high levels in mature homogeneous populations of either SMA+ or GFAP+ cells. To Schwann cells (Jessen and Mirsky, 2002), was also detectable further characterize the glia cells, high-density cultures treated at low levels in our cultures by immunohistochemistry and RT- with FGF2 and BMP2 were stained for both GFAP and PCR (data not shown). MYP0 protein was found in a high p75NGFR. A majority of cells were GFAP+/p75NGFR+, consistent proportion of cells but at low levels (data not shown), similar to immature neural crest cells (Hagedorn et al., 1999). These cells exhibit characteristics indicating a non-myelinating, non- CNS glia identity and are also consistent with the cell types seen in the CPm. BMP-exposure did not reproducibly generate cells that expressed peripheral neuronal markers. We addressed this by adding retinoic acid, a known posteriorizing factor in neural crest induction (Villanueva et al., 2002), during the FGF2- BMP2 co-treatment in acutely dissociated cells. We then included B27, NGF, Heregulin-␤1, BDNF and GDNF along with BMP2 during the FGF2-withdrawal phase. We observed cells expressing peripherin (Fig. 6G-I), an intermediate filament found in peripheral neurons and some population of CNS neurons projecting into the periphery (e.g. cranial nerve ganglia and motoneurons), but also in rare neuronal populations of the neocortex (Rhrich-Haddout et al., 1997). Nearly all (92%) of the peripherin+ cells co-stained for Brn3a (Brn3.0, Fig. 6H), a POU transcription factor that is expressed in nearly all dorsal root sensory neurons (Anderson, 1999), but not in the cortex. We could not reproducibly find cells expressing tyrosine hydoxylase (TH) (data not shown), an enzyme expressed in peripheral autonomic neurons. Thus, Journal of Cell Science under the appropriate posteriorizing conditions we were able to identify cells consistent with a peripheral neuronal identity, although the absence of sufficient trophic factors might account for the low frequency. We were unable to identify other neural crest derivatives such as melanoblasts, melanocytes or chondroblasts as measured by Kit, MITF, Trp2 or collagen 2␣1 protein and Trp2 and collagen 2␣1 mRNA (not shown). In contrast to FGF2-BMP2 co-treated cells, the cultures grown only in FGF2 (with or without retinoic acid) did not Fig. 6. BMP-treated CNS stem cells differentiate to non-myelinating, generate peripherin+ cells (Fig. 6I), GFAP+/p75NGFR+ cells non-CNS glia. (A-F) Generation of glia cells using paradigm 1 after (Fig. 6C) or GalC+/p75NGFR+ cells (Fig. 6F), nor cells staining high-density plating. (A-C) Control cultures generate immature GFAP+ (green) astrocytes (A), whereas BMP2 co-treatment yields for more mature Schwann cell markers such as MAG, MBP or distinctively flattened cells, most co-expressing GFAP (green) and MYP0 (data not shown). Thus, the generation of these non- p75NGFR (red), consistent with a non-myelinating Schwann cell CNS differentiated derivatives depended on the specific actions phenotype (B, quantitation in C). (D-F) Control cultures generate of BMPs. only small numbers of GalC+ CNS oligodendrocytes and no GalC+/p75NGFR+ cells (D), wheras BMP2 co-treatment yields morphologically distinct GalC+ (red) and 75NGFR+ (green) co- Both fetal and adult CNS stem cells generate expressing cells (D, quantitation in F). (G-I) Peripherin+ neurons dorsalized derivatives could be generated in acute cultures only by adding retinoic acid The data from E14.5 cortical cells showed that FGF2 conferred during a 3-day FGF2/BMP2 co-treatment, followed by BDNF, a competence to generate mesenchymal derivatives similar to GDNF, NGF and HRG during a 7-day mitogen withdrawal. Peripherin+ cells had long and sometimes branched processes (G, CPm or neural crest in response to BMPs. Adult CNS stem quantitation in I); 92% of peripherin+ cells co-expressed Brn3a (H), cells are similar to fetal stem cells in their responses to factors consistent with a peripheral neuron identity. Graphs show mean ± that control neuronal and glial differentiation (Johe et al., s.e.m. (n=3-4). Bars, 20 ␮m (A,B); 10 ␮m (D,E); 80 ␮m (G), 10 ␮m 1996), indicating that they share common signaling (H). mechanisms. BMP2 and BMP4 are expressed in the adult BMP2 and FGF2 in CNS stem cells 5857 30% at day 3 and decreased thereafter (Fig. 7E), consistent with the idea that p75NGFR expression often marks a transient precursor state (Morrison et al., 1999). Whereas control cultures generated no SMA+ cells (Fig. 8A), FGF2/BMP2 co-treatment generated cells that expressed both SMA (8.1±2.7%, Fig. 8B,C) and calponin (8.8±3.7%). Interestingly, the proportion of smooth-muscle cells (SMA+) induced by BMP exposure decreased, depending on the isolation age. When controlling for culture conditions, the percentage of SMA+ cells progressively decreased from E14.5 to E18.5 to adult cells (Fig. 8C). By contrast, adult cells generated larger numbers of glia cells in response to BMP2. Whereas differentiated control cultures contained weakly stained GFAP+ cells with an immature astrocytic morphology (Fig. 8D), BMP2 co-treatment generated large numbers of distinctly flattened, strongly- expressing GFAP+ cells (Fig. 8E), most of which co-expressed p75NGFR (not shown). The proportion of total GFAP+ cells increased with the age of the donor animal (Fig. 8F). A smaller proportion of BMP2-treated adult cells were GalC+/p75NGFR+ (5.8±3.3%) although this also represented an increase over that seen in fetal cultures (Fig. 8G-I). Since EGF is a commonly used mitogen for adult SVZ stem cells, we tested EGF-BMP2 co-treatment and found neither p75NGFR+ nor SMA+ cells in these cultures (data not shown). Thus, adult SVZ cells also retain the capacity to be dorsalized to CPm or neural crest derivatives in response to BMP2 and also require FGF2 for this Fig. 7. FGF2 and BMP2 induce adult rat SVZ stem cells to dorsalized precursors. Adult SVZ stem cells were expanded in FGF2 action. Whereas BMP-mediated dorsalization occurs in CNS with noggin prior to passage. (A) RT-PCR time course analysis of stem cells from different ages, the propensity toward glial Snail1 and Snail2 expression in passage 1, during FGF2 ± BMP2 differentiation is higher in adult compared with fetal CNS stem exposure. GAPDH expression is shown in reverse-transcribed (+) cells. and non-transcribed (–) GAPDH samples as loading controls. (B-E) Comparison of p75NGFR expression after low-density plating; clones marked after 5 days expansion (116±54 cells/clone) were Discussion further expanded with FGF2 ± BMP2. Expression of p75NGFR is The neural crest contains multipotent cells that can acquire

Journal of Cell Science absent in (B) control cultures but prevalent in (C) BMP2 co-treated NGFR+ PNS and other fates (Baroffio et al., 1988; Bronner-Fraser and cells. Percentage of clones containing (D) at least one p75 cell Fraser, 1988). Many studies have identified roles for FGF, and total percentage of (E) p75NGFR+ cells during FGF2 expansion without (ᮀ) or with (᭜) BMP2. Graphs show mean ± s.e.m. (n=2). BMP and Wnt signals in the induction of neural crest from the Bars 20 ␮m. neural-plate-stage embryo (Aybar et al., 2002; Garcia-Castro et al., 2002; LaBonne and Bronner-Fraser, 1998; Villanueva et al., 2002). Specifically, a BMP-dependent step precedes the mouse subventricular zone (SVZ) and their gliogenic actions migration of neural crest from the neural tube. The on multipotent SVZ cells are antagonized by the secreted factor transcription factors Snail1 and Snail2 are known to regulate noggin (Lim et al., 2000). We added noggin to our SVZ the EMT that is required for neural crest migration. We and cultures during expansion to minimize any confounding effect others have found evidence that, CNS stem cells, which of endogenous BMPs. After the first passage, we tested adult typically give rise to neuronal and glial fates (Johe et al., 1996), SVZ cultures for mesenchymal induction. can be dorsally induced by BMPs to generate cells with neural BMP2 treatment upregulated the expression of Snail1 and crest characteristics (Alexanian and Sieber-Blum, 2003; Snail2 beginning at day 1 and continuing through 5 days of Gajavelli et al., 2004; Mujtaba et al., 1998; Panchision et al., treatment as measured by RT-PCR (Fig. 7A). The neural crest 2001; Rajan et al., 2003). In this study, we show that fetal markers AP-2, noelin-1 and foxd3 were induced after 4 hours telencephalic and adult CNS stem cells respond only to the of treatment, whereas the peripheral glia marker MYP0 was combination of BMP2 and FGF2 by inducing Snail1 and induced after 5 days (not shown). In clonal analysis, no Snail2, migrating from cortical explants, assuming a p75NGFR+ p75NGFR+ cells were seen in FGF2-expanded cells (Fig. 7B), mesenchymal morphology and efficiently differentiating into whereas addition of BMP2 led to the induction of p75NGFR+, smooth-muscle and non-CNS glia cells. starting at day 1 and peaking at day 3 of the treatment (Fig. This finding is surprising for a number of reasons. First, key 7C-E). Expression of p75NGFR occurred faster in adult stem regional-identity regulators such as Foxg1 (Bf1), Nkx-2.1, Otx- cells than in the fetal stem cells; almost all cell clusters 1, Emx-2 and members of the Dlx gene family are expressed contained p75NGFR+ cells after 2 days of BMP2 co-treatment in the forebrain by E9.5 (Rubenstein et al., 1998), well before (Fig. 7D). In contrast to the fetal cultures, the maximum the ages we tested. Second, the neural crest is distinct from the frequency of p75NGFR expression in adult cultures was only CNS at cranial levels by approximately E9.5 (Serbedzija et al., 5858 Journal of Cell Science 118 (24) 1992), at which point neural crest precursors have migrated out crest of the posterior neural tube is transient; the last neural of the neuroepithelium and into the surrounding mesenchyme. crest cells leave the neural tube at E9.5 in the mouse Third, fate-mapping of chicken cells shows that neural crest (Serbedzija et al., 1992), which corresponds in chicken does not arise from the telencephalic neuroepithelium during development to the Hamburger and Hamilton (HH) stage 15 normal development, but instead arises from regions caudal to (Hamburger and Hamilton, 1992). However, studies in chicken the epiphysis of the mid-diencephalon (Couly and Le Douarin, show that expression of Snail and Slug in the CPm occur at HH 1987; Le Douarin and Kalcheim, 1999). stages 21-41 (Marin and Nieto, 2004), corresponding to However, the BMP-generated cells in our study have approximately E12.5-18.5 in mouse gestation and beyond the properties that are most similar to CPm, a cell type in close time that neural crest emigrates from the neural tube. Chick- proximity to both CPe, a forebrain cell type, and cranial quail grafting experiments did not identify a contribution from mesenchyme, a neural crest derivative. Both the migrating either CPe (Wilting and Christ, 1989) or neural crest (Le neural crest and CPm share expression of the Snail Douarin and Kalcheim, 1999) to the CPm, leaving its transcription factors (Marin and Nieto, 2004). Snail1 and developmental origins unclear. However, our results support Snail2 are centrally involved in cell movement, in particular the possibility that CPm might derive from anterior CNS cells. during the epithelial to mesenchymal transition (EMT) of cells Alternatively, grafting and genetic manipulation experiments before migration (Barrallo-Gimeno and Nieto, 2005; De suggest that CNS precursors can adopt new fates that are Craene et al., 2005). The expression of Snail genes in the neural appropriate to their ectopic location in the CNS or PNS (Brustle et al., 1995; Hitoshi et al., 2002; Korade and Frank, 1996). Extensive dorsalization of cortex to Foxj1+ CPe, the most dorsal CNS cell type in the forebrain, occurs in transgenic mice expressing a constitutively active BMP receptor IA (c.a.BMPR-IA). Whereas CPe was the principal forebrain cell type generated in these animals, FGF2-expanded cortical stem cells transfected with the same c.a.BMPR-IA constructs very robustly generated p75NGFR+ precursors and smooth muscle (Panchision et al., 2001). Our current studies explain the difference between the transgenic and in-vitro data by showing a specific requirement for FGF2 signaling in the dorsalization of CNS stem cells to mesenchymal cell types. This suggests that the levels of FGF2 is limiting in this dorsal conversion in vivo. FGF signaling is crucial for supporting CNS development in vivo (Delaune et al., 2004; Rubenstein, 2000; Vaccarino et al., 1999). Journal of Cell Science Since FGF2 is most frequently used as an obligatory mitogen during in-vitro stem cell expansion (Panchision and McKay, 2002), it is difficult to measure its interactions with other factors that regulate stem cell fate. Our explant culture assay shows that FGF2 is required for EMT and a neural- crest-like fate and that its actions are not duplicated by EGF or IGF1, even though these factors also promote proliferation. FGF signaling has recently been implicated in the ventralization of neural stem cells and neural explants (Gabay et al., 2003; Kessaris et al., 2004; Kuschel et al., 2003). FGF2 expansion of cortical Fig. 8. BMP-treated adult SVZ stem cells preferentially differentiate into non- stem cells was shown to increase the frequency of CNS glia. Passaged adult SVZ stem cells were plated at low density, FGF2- Olig2+ cells, which mark ventral precursor cells that expanded for 3 days and then treated with FGF2 ± BMP2 for 7 days. are capable of generating oligodendrocytes (Gabay (A) Control cultures generate no SMA+ cells; (B) BMP2 co-treatment yields et al., 2003). Whereas most oligodendrocyte flattened SMA+ cells; (C) Percent SMA+ cells is decreased in cells from older generation is hedgehog-dependent (Kessaris et al., embyos and adults. (D) Control cultures generate thin, immature GFAP+ cells; + + 2004; Tekki-Kessaris et al., 2001), an FGF2- (E) BMP2 co-treatment yields flattened GFAP cells; (F) Percent GFAP cells dependent and hedgehog-independent pathway was increases as a function of animal age (F). (G-I) FGF2-expanded cultures and also identified, suggesting that FGF2 acts in part by cultures where FGF2 was withdrawn generate infrequent GalC+ cells with immature CNS oligodendrocyte morphologies and do not contain p75NGFR+ antagonizing BMP signaling (Chandran et al., 2003). cells (G), whereas BMP2 co-treatment yields flattened p75NGFR+ cells, some of Our results indicate that FGF2, rather than which co-express GalC (H). The percentage of FGF2-BMP2-generated antagonizing dorsalization, is actually required for GalC+/p75NGFR+ cells increases as a function of animal age (I). All graphs show BMP2-mediated dorsalization. Thus, FGF2 does not mean ± s.e.m., n=3-4. Bar, 20 ␮m. appear to be ventralizing but rather acts as a BMP2 and FGF2 in CNS stem cells 5859

permissive signal (Freeman and Gurdon, 2002) that allows specification of vertebrate primary sensory neurons. Curr. Opin. stem cells to respond to other instructive cues. This idea is Neurobiol. 9, 517-524. consistent with the requirement for FGF2 in such diverse Aybar, M. J., Glavic, A. and Mayor, R. (2002). Extracellular signals, cell ␤ interactions and transcription factors involved in the induction of the neural responses like Wnt- and -catenin-mediated precursor crest cells. Biol. Res. 35, 267-275. proliferation (Israsena et al., 2004), hedgehog-mediated Aybar, M. J., Nieto, M. A. and Mayor, R. (2003). Snail precedes slug in the oligodendrocyte generation (Kessaris et al., 2004), induction of genetic cascade required for the specification and migration of the Xenopus neural crest by paraxial mesoderm (Monsoro-Burq et al., 2003) neural crest. Development 130, 483-494. and BMP-mediated dorsalization of CNS precursors to neural- Backman, M., Machon, O., Mygland, L., van den Bout, C. J., Zhong, W., Taketo, M. M. and Krauss, S. (2005). Effects of canonical Wnt signaling crest-like cells (Gajavelli et al., 2004; Mujtaba et al., 1998; on dorso-ventral specification of the mouse telencephalon. Dev. Biol. 279, Panchision et al., 2001; Rajan et al., 2003). 155-168. We found that CNS stem cells from tissue of older embryos Baroffio, A., Dupin, E. and Le Douarin, N. M. (1988). Clone-forming ability gave rise to increasing proportions of glia cells at the expense of and differentiation potential of migratory neural crest cells. Proc. Natl. Acad. Sci. USA 85, 5325-5329. smooth muscle and neurons. At E14.5, glial cells were generated Barrallo-Gimeno, A. and Nieto, M. A. (2005). The Snail genes as inducers only in regions of high local density in response to BMP of cell movement and survival: implications in development and cancer. treatment. By contrast, adult stem cells differentiated more Development 132, 3151-3161. uniformly to non-CNS glia. Previous studies show that these Beattie, M. S., Harrington, A. W., Lee, R., Kim, J. Y., Boyce, S. L., Longo, responses mirror the in-vivo developmental progression of cell- F. M., Bresnahan, J. C., Hempstead, B. L. and Yoon, S. O. (2002). ProNGF induces p75-mediated death of oligodendrocytes following spinal type differentiation (Panchision and McKay, 2002). Increased cord injury. Neuron 36, 375-386. apoptosis followed by neuronal differentiation occurs in Bronner-Fraser, M. (2002). Molecular analysis of neural crest formation. J. response to BMP receptor IB (BMPR-IB) activation in Physiol. Paris 96, 3-8. transgenic embryos (Panchision et al., 2001). Cultured neural Bronner-Fraser, M. and Fraser, S. E. (1988). Cell lineage analysis reveals precursors isolated from progressively older embryos also multipotency of some avian neural crest cells. Nature 335, 161-164. Brustle, O., Maskos, U. and McKay, R. D. (1995). Host-guided migration respond to BMPs with first apoptosis (Furuta et al., 1997; allows targeted introduction of neurons into the embryonic brain. Neuron Graham et al., 1996), then neuronal differentiation in mid- 15, 1275-1285. gestation (Li et al., 1998; Mehler et al., 2000) and finally with Chandran, S., Kato, H., Gerreli, D., Compston, A., Svendsen, C. N. and glial differentiation in late-gestation and adulthood (Gross et al., Allen, N. D. (2003). FGF-dependent generation of oligodendrocytes by a 1996; Mabie et al., 1999). However, BMP treatment of EGF- hedgehog-independent pathway. Development 130, 6599-6609. Couly, G. F. and Le Douarin, N. M. (1987). Mapping of the early neural expanded precursors generates glia cells with a CNS astrocyte- primordium in quail-chick chimeras. II. The prosencephalic neural plate and like morphology (Gross et al., 1996). We show that BMP neural folds: implications for the genesis of cephalic human congenital treatment of FGF2-expanded precursors generates glia cells with abnormalities. Dev. Biol. 120, 198-214. non-CNS characteristics. An age-progressive bias in fate choice De Craene, B., van Roy, F. and Berx, G. (2005). Unraveling signalling cascades for the Snail family of transcription factors. Cell Signal 17, 535-547. is also seen in PNS stem cells (Kruger et al., 2002; White et al., Delaune, E., Lemaire, P. and Kodjabachian, L. (2004). Neural induction in 2001). Our data suggest that the same mechanisms control the Xenopus requires early FGF signalling in addition to BMP inhibition. changing developmental bias for both CNS and non-CNS fates. Development 132, 299-310. In conclusion, we show that both fetal and adult cortical stem Dono, R. (2003). Fibroblast growth factors as regulators of central nervous Journal of Cell Science cells have the capacity to generate neural-crest-like cells and that system development and function. Am. J. Physiol. Regul. Integr. Comp. Physiol. 284, R867-R881. the age of the stem cells biases the type of derivatives that are Drago, J., Murphy, M., Carroll, S. M., Harvey, R. P. and Bartlett, P. F. generated. FGF2 is crucial for permitting this BMP-mediated (1991). Fibroblast growth factor-mediated proliferation of central nervous response. Since many secreted molecules exert both morphogenic system precursors depends on endogenous production of insulin-like growth and mitogenic actions (Panchision and McKay, 2002), it will be factor I. Proc. Natl. Acad. Sci. USA 88, 2199-2203. of interest to determine whether these actions are mechanistically Etchevers, H. C., Couly, G. and Le Douarin, N. M. (2002). Morphogenesis of the branchial vascular sector. Trends Cardiovasc. Med. 12, 299-304. distinct. The ability of BMPs and FGFs to rapidly and efficiently Freeman, M. and Gurdon, J. B. (2002). Regulatory principles of re-direct the differentiation of CNS stem cells will allow further developmental signaling. Annu. Rev. Cell Dev. Biol. 18, 515-539. studies to define the mechanism of this interaction. Furuta, Y., Piston, D. W. and Hogan, B. L. (1997). Bone morphogenetic proteins (BMPs) as regulators of dorsal forebrain development. We thank Story Landis, Vittorio Gallo, Tarik Haydar, Raji Development 124, 2203-2212. Padmanabhan, Raja Kittappa, Rea Ravin, Franziska Sailer and Gabay, L., Lowell, S., Rubin, L. L. and Anderson, D. J. (2003). Andreas Androutsellis-Theotokis for advice and discussion, and Deregulation of dorsoventral patterning by FGF confers trilineage differentiation capacity on CNS stem cells in vitro. Neuron 40, 485-499. Bechien Wu, Peter Warinner and Kimber Bogush for technical Gajavelli, S., Wood, P. M., Pennica, D., Whittemore, S. R. and Tsoulfas, assistance. BMP ligands in initial experiments were provided by P. (2004). BMP signaling initiates a neural crest differentiation program in Wyeth Pharmaceuticals. We are grateful to Robert Adelstein (National embryonic rat CNS stem cells. Exp. Neurol. 188, 205-223. Heart, Lung and Blood Institute, NIH, MD), Eric Turner (University Gammill, L. S. and Bronner-Fraser, M. (2003). Neural crest specification: of California, San Diego, CA) and Juan Archelos (University Hospital migrating into genomics. Nat. Rev. Neurosci. 4, 795-805. Graz, Austria, Europe) for their gifts of antibodies. M.H.M.S. was Garcia-Castro, M. I., Marcelle, C. and Bronner-Fraser, M. (2002). supported by a grant from the Swiss National Science Foundation Ectodermal Wnt function as a neural crest inducer. Science 297, 848-851. (SNSF 3135-54876.98). This research was supported in part by the Graham, A., Koentges, G. and Lumsden, A. (1996). Neural crest apoptosis Intramural Research Program of the NIH, NINDS. and the establishment of craniofacial pattern: an honorable death. Mol. Cell Neurosci. 8, 76-83. Gross, R. E., Mehler, M. F., Mabie, P. C., Zang, Z., Santschi, L. and Kessler, J. A. (1996). Bone morphogenetic proteins promote astroglial References lineage commitment by mammalian subventricular zone progenitor cells. Alexanian, A. R. and Sieber-Blum, M. (2003). Differentiating adult Neuron 17, 595-606. hippocampal stem cells into neural crest derivatives. Neuroscience 118, 1-5. Grove, E. A. and Fukuchi-Shimogori, T. (2003). Generating the cerebral Anderson, D. J. (1999). Lineages and transcription factors in the cortical area map. Annu. Rev. Neurosci. 26, 355-380. 5860 Journal of Cell Science 118 (24)

Gunhaga, L., Marklund, M., Sjodal, M., Hsieh, J. C., Jessell, T. M. and Monsoro-Burq, A. H., Fletcher, R. B. and Harland, R. M. (2003). Neural Edlund, T. (2003). Specification of dorsal telencephalic character by crest induction by paraxial mesoderm in Xenopus embryos requires FGF sequential Wnt and FGF signaling. Nat. Neurosci. 6, 701-707. signals. Development 130, 3111-3124. Hagedorn, L., Suter, U. and Sommer, L. (1999). P0 and PMP22 mark a Moon, R. T., Kohn, A. D., De Ferrari, G. V. and Kaykas, A. (2004). WNT multipotent neural crest-derived cell type that displays community effects and beta-catenin signalling: diseases and therapies. Nat. Rev. Genet. 5, 691- in response to TGF-beta family factors. Development 126, 3781-3794. 701. Hamburger, V. and Hamilton, H. L. (1992). A series of normal stages in the Morrison, S. J., White, P. M., Zock, C. and Anderson, D. J. (1999). development of the chick embryo. 1951. Dev Dyn 195, 231-272. Prospective identification, isolation by flow cytometry, and in vivo self- Hebert, J. M., Mishina, Y. and McConnell, S. K. (2002). BMP signaling is renewal of multipotent mammalian neural crest stem cells. Cell 96, 737-749. required locally to pattern the dorsal telencephalic midline. Neuron 35, Mujtaba, T., Mayer-Proschel, M. and Rao, M. S. (1998). A common neural 1029-1041. progenitor for the CNS and PNS. Dev. Biol. 200, 1-15. Hitoshi, S., Tropepe, V., Ekker, M. and van der Kooy, D. (2002). Neural Nieto, M. A. (2002). The snail superfamily of zinc-finger transcription factors. stem cell lineages are regionally specified, but not committed, within distinct Nat. Rev. Mol. Cell. Biol. 3, 155-166. compartments of the developing brain. Development 129, 233-244. Panchision, D. M. and McKay, R. D. (2002). The control of neural stem cells Israsena, N., Hu, M., Fu, W., Kan, L. and Kessler, J. A. (2004). The by morphogenic signals. Curr. Opin. Genet. Dev. 12, 478-487. presence of FGF2 signaling determines whether beta-catenin exerts effects Panchision, D. M., Pickel, J. M., Studer, L., Lee, S. H., Turner, P. A., Hazel, on proliferation or neuronal differentiation of neural stem cells. Dev. Biol. T. G. and McKay, R. D. (2001). Sequential actions of BMP receptors 268, 220-231. control neural precursor cell production and fate. Genes Dev. 15, 2094-2110. Jessen, K. R. and Mirsky, R. (2002). Signals that determine Schwann cell Rajan, P., Panchision, D. M., Newell, L. F. and McKay, R. D. (2003). BMPs identity. J. Anat. 200, 367-376. signal alternately through a SMAD or FRAP-STAT pathway to regulate fate Johe, K. K., Hazel, T. G., Muller, T., Dugich-Djordjevic, M. M. and choice in CNS stem cells. J. Cell Biol. 161, 911-921. McKay, R. D. (1996). Single factors direct the differentiation of stem cells Rhrich-Haddout, F., Klosen, P., Portier, M. M. and Horvat, J. C. (1997). from the fetal and adult central nervous system. Genes Dev. 10, 3129-3140. Expression of peripherin, NADPH-diaphorase and NOS in the adult rat Kessaris, N., Jamen, F., Rubin, L. L. and Richardson, W. D. (2004). neocortex. NeuroReport 8, 3313-3316. Cooperation between sonic hedgehog and fibroblast growth factor/MAPK Rubenstein, J. L. (2000). Intrinsic and extrinsic control of cortical signalling pathways in neocortical precursors. Development 131, 1289- development. Novartis Found. Symp. 228, 67-82; 109-113. 1298. Rubenstein, J. L., Shimamura, K., Martinez, S. and Puelles, L. (1998). Kim, J.-H., Panchision, D. M., Kittappa, R. and McKay, R. D. (2003). Regionalization of the prosencephalic neural plate. Annu. Rev. Neurosci. 21, Generating CNS neurons from embryonic, fetal and adult stem cells. 445-477. Differentiation of Embryonic Stem Cells. New York: Academic Press. Saint-Jeannet, J. P., He, X., Varmus, H. E. and Dawid, I. B. (1997). Kleber, M., Lee, H. Y., Wurdak, H., Buchstaller, J., Riccomagno, M. M., Regulation of dorsal fate in the neuraxis by Wnt-1 and Wnt-3a. Proc. Natl. Ittner, L. M., Suter, U., Epstein, D. J. and Sommer, L. (2005). Neural Acad. Sci. USA 94, 13713-13718. crest stem cell maintenance by combinatorial Wnt and BMP signaling. J. Serbedzija, G. N., Bronner-Fraser, M. and Fraser, S. E. (1992). Vital dye Cell Biol. 169, 309-320. analysis of cranial neural crest cell migration in the mouse embryo. Knecht, A. K. and Bronner-Fraser, M. (2002). Induction of the neural crest: Development 116, 297-307. a multigene process. Nat. Rev. Genet. 3, 453-461. Shah, N. M., Groves, A. K. and Anderson, D. J. (1996). Alternative neural Korade, Z. and Frank, E. (1996). Restriction in cell fates of developing spinal crest cell fates are instructively promoted by TGFbeta superfamily members. cord cells transplanted to neural crest pathways. J. Neurosci. 16, 7638-7648. Cell 85, 331-343. Kruger, G. M., Mosher, J. T., Bixby, S., Joseph, N., Iwashita, T. and Tekki-Kessaris, N., Woodruff, R., Hall, A. C., Gaffield, W., Kimura, S., Morrison, S. J. (2002). Neural crest stem cells persist in the adult gut but Stiles, C. D., Rowitch, D. H. and Richardson, W. D. (2001). Hedgehog- undergo changes in self-renewal, neuronal subtype potential, and factor dependent oligodendrocyte lineage specification in the telencephalon. responsiveness. Neuron 35, 657-669. Development 128, 2545-2554. Kuschel, S., Ruther, U. and Theil, T. (2003). A disrupted balance between Trainor, P. A., Sobieszczuk, D., Wilkinson, D. and Krumlauf, R. (2002). Journal of Cell Science Bmp/Wnt and Fgf signaling underlies the ventralization of the Gli3 mutant Signalling between the hindbrain and paraxial tissues dictates neural crest telencephalon. Dev. Biol. 260, 484-495. migration pathways. Development 129, 433-442. LaBonne, C. and Bronner-Fraser, M. (1998). Neural crest induction in Tsai, R. Y. and McKay, R. D. (2000). Cell contact regulates fate choice by Xenopus: evidence for a two-signal model. Development 125, 2403-2414. cortical stem cells. J. Neurosci. 20, 3725-3735. Le Douarin, N. M. and Kalcheim, C. (1999). The Neural Crest. New York: Vaccarino, F. M., Schwartz, M. L., Raballo, R., Nilsen, J., Rhee, J., Zhou, Cambridge University Press. M., Doetschman, T., Coffin, J. D., Wyland, J. J. and Hung, Y. T. (1999). Lee, H. Y., Kleber, M., Hari, L., Brault, V., Suter, U., Taketo, M. M., Changes in cerebral cortex size are governed by fibroblast growth factor Kemler, R. and Sommer, L. (2004). Instructive role of Wnt/beta-catenin in during embryogenesis. Nat. Neurosci. 2, 246-253. sensory fate specification in neural crest stem cells. Science 303, 1020-1023. Vicario-Abejon, C., Collin, C., Tsoulfas, P. and McKay, R. D. (2000). Lee, K. J. and Jessell, T. M. (1999a). The specification of dorsal cell fates in Hippocampal stem cells differentiate into excitatory and inhibitory neurons. the vertebrate central nervous system. Annu. Rev. Neurosci. 22, 261-294. Eur. J. Neurosci. 12, 677-688. Lee, K. J. and Jessell, T. M. (1999b). The specification of dorsal cell fates in Villanueva, S., Glavic, A., Ruiz, P. and Mayor, R. (2002). Posteriorization the vertebrate central nervous system. Annu. Rev. Neurosci. 22, 261-294. by FGF, Wnt, and retinoic acid is required for neural crest induction. Dev. Li, W., Cogswell, C. A. and LoTurco, J. J. (1998). Neuronal differentiation Biol. 241, 289-301. of precursors in the neocortical ventricular zone is triggered by BMP. J. White, P. M., Morrison, S. J., Orimoto, K., Kubu, C. J., Verdi, J. M. and Neurosci. 18, 8853-8862. Anderson, D. J. (2001). Neural crest stem cells undergo cell-intrinsic Lim, D. A., Tramontin, A. D., Trevejo, J. M., Herrera, D. G., Garcia- developmental changes in sensitivity to instructive differentiation signals. Verdugo, J. M. and Alvarez-Buylla, A. (2000). Noggin antagonizes BMP Neuron 29, 57-71. signaling to create a niche for adult neurogenesis. Neuron 28, 713-726. Wilson, S. W. and Rubenstein, J. L. (2000). Induction and dorsoventral Mabie, P. C., Mehler, M. F. and Kessler, J. A. (1999). Multiple roles of bone patterning of the telencephalon. Neuron 28, 641-651. morphogenetic protein signaling in the regulation of cortical cell number Wilting, J. and Christ, B. (1989). An experimental and ultrastructural study and phenotype. J. Neurosci. 19, 7077-7088. on the development of the avian choroid plexus. Cell Tissue Res. 255, 487- Marin, F. and Nieto, M. A. (2004). Expression of chicken slug and snail in 494. mesenchymal components of the developing central nervous system. Dev. Wu, J., Saint-Jeannet, J. P. and Klein, P. S. (2003). Wnt-frizzled signaling Dyn. 230, 144-148. in neural crest formation. Trends Neurosci. 26, 40-45. Mehler, M. F., Mabie, P. C., Zhu, G., Gokhan, S. and Kessler, J. A. (2000). Zorick, T. S. and Lemke, G. (1996). Schwann cell differentiation. Curr. Opin. Developmental changes in progenitor cell responsiveness to bone Cell Biol. 8, 870-876. morphogenetic proteins differentially modulate progressive CNS lineage Zorick, T. S., Syroid, D. E., Arroyo, E., Scherer, S. S. and Lemke, G. fate. Dev. Neurosci. 22, 74-85. (1996). The Transcription Factors SCIP and Krox-20 Mark Distinct Stages Meulemans, D. and Bronner-Fraser, M. (2004). Gene-regulatory and Cell Fates in Schwann Cell Differentiation. Mol. Cell Neurosci. 8, 129- interactions in neural crest evolution and development. Dev. Cell 7, 291-299. 145.