Neural induction and factors that stabilize a neural fate

Abstract

The neural ectoderm of vertebrates forms when the BMP signaling pathway is suppressed. Herein we review the molecules that directly antagonize extracellular BMP and the signaling pathways that further contribute to reduce BMP activity in the neural ectoderm. Downstream of neural induction, a large number of “neural fate stabilizing” (NFS) transcription factors are expressed in the presumptive neural ectoderm, developing neural tube, and ultimately in neural stem cells. Herein we review what is known about their activities during normal development to maintain a neural fate and regulate neural differentiation. Further elucidation of how the NFS genes interact to regulate neural specification and differentiation should ultimately prove useful for regulating the expansion and differentiation of neural stem and progenitor cells.

Introduction

Nearly 100 years ago, Hilde Mangold and Hans Spemann (1924) excised small bits of tissue from the point at which cells first involute in the amphibian gastrula and grafted these into the ventral belly region of host embryos. Surprisingly, a secondary body axis formed at the transplant site, and it included a neural tube and axial mesoderm (notochord, somites). By using explants and hosts from differently pigmented species so that the donor cells could be identified, they demonstrated that the transplanted piece, now known as Spemann’s “organizer”, did not give rise to neural ectoderm, but instead formed dorsal mesoderm. They reasoned that the transplanted cells signaled adjacent host cells to become the neural tissue of the secondary axis; thus, the phenomenon of neural induction was discovered. Equivalent organizing regions with neural inducing capacity are found in zebrafish (shield), chick (Henson’s node), and mouse (node) embryos (Beddington, 1994; Shih and Fraser, 1996; Smith and Schoenwolf, 1998; Boettger et al., 2001). Hence, neural induction is one of the primary steps in vertebrate neural development, and understanding how it is accomplished is of fundamental importance.

Since this landmark observation, numerous studies focused on identifying the inducing molecule(s) that are necessary to establish neural fate in vertebrate embryos. In addition, a large number of transcription factors whose expression is initiated by neural inductive signaling was identified. How these factors cooperatively stabilize the neural fate of embryonic cells is only beginning to be appreciated. While understanding the acquisition of neural fate during embryogenesis is exciting in its own right, it also appears that these same signaling factors and transcriptional regulators may be expressed in endogenous niches of neural stem cells in the adult brain. Therefore, the developmental program that specifies a neural fate in the embryo may prove to be critically important for coaxing neural stem cells to repopulate damaged tissue. This review will summarize the current state of knowledge of the process of neural induction, the genes involved in neural fate stabilization, and speculate on how this information may play a role in regulating the differentiation of neural stem cells.

Neural Induction

The organizer forms in the dorsal mesoderm in response to secreted factors from a signaling center located in the adjacent presumptive dorsal endoderm, which in amphibians is called the Nieuwkoop Center (Nieuwkoop, 1967; Gerhart et al., 1989; Agius et al., 2000; Vonica and Gumbiner, 2007). In Xenopus, the Nieuwkoop Center forms where the expression domains of VegT, Vg1, Nodal-related (Nr) proteins, and β-Catenin overlap (Fig. 1). The Spemann-Mangold experiment predicted that a neural inducing substance emanated from the organizer. While several studies sought to identify the neural inducer(s), more recent work led to the hypothesis that ectoderm forms neural tissue not by an activating signal, but by default (Sato and Sargent, 1989; Wilson and Hemmati-Brivanlou, 1995; Hawley et al., 1995; Weinstein and Hemmati-Brivanlou, 1997). Pieces of naïve embryonic ectoderm dissected from the animal pole of Xenopus blastulae (called animal cap [AC] ectodermal explants) form non-neural epidermis when cultured without the addition of growth factors (Fig. 2). However, if the AC ectodermal explant is first dissociated into single cells, the cells become neural, indicating that: (1) short range cell-to-cell signaling is required for the formation of epidermis, and (2) in the absence of this signaling, neural tissue forms by default (Sato and Sargent, 1989; Wilson and Hemmati-Brivanlou, 1995). In support of this hypothesis, experiments showed that the inhibition of transforming growth factor-β (TGF-β) signaling through the use of a dominant-negative Activin receptor caused AC ectodermal explants to express neural marker genes (Hemmati-Brivanlou et al., 1992; Hemmati-Brivanlou and Melton, 1994; Hemmati-Brivanlou and Thomsen, 1995; Hawley et al., 1995). However, the dominant-negative Activin receptor interferes with signaling of all of the members of the TGF-β family, including TGF-β, Activin, Nr, and bone morphogenetic proteins (BMPs). Further work demonstrated that BMPs are the important ligands for neural versus epidermal fate induction.

Figure 1.

The dorsal-ventral axis is specified by the Nieuwkoop center and the Spemann organizer in Xenopus. (A) Prior to gastrulation the dorsal axis is specified by the Nieuwkoop center (green circle), formed as a result of overlapping expression of VegT, Vg1, Nodals (Xnr), and β-Catenin (βcat). (B) The Nieuwkoop center secretes dorsalizing molecules that induce the formation of the Spemann organizer (blue oval) in the dorsal mesoderm. The organizer expresses Chordin (Chd), Noggin (Nog), Follistatin (Foll), Cerberus (Cer), and in Xenopus, Nodal related-3 (Xnr3). These molecules actively antagonize BMP, which is secreted from the ventral side (red arrows). V is ventral; D is dorsal. Line intersecting the Nieuwkoop center represents the dorsal blastopore lip.

Figure 2.

The default model of neural induction in Xenopus. An ectodermal explant dissected from the animal cap of a blastula stage embryo forms epidermis (light blue). If BMP signaling is inhibited in those cells by expressing a dominant-negative Activin receptor (dnActR) or the extracellular antagonist Noggin, then neural tissue forms (dark blue). Dissociation of the explant into single cells without the addition of exogenous factors also leads to the formation of neural tissues. However, if exogenous BMP is added to the dispersed cells, they adopt an epidermal fate (light blue).

BMP signaling

Several studies pointed to the BMP signaling pathway as key for neural induction, but as a negative regulator, not a positive inducer. First, using expression screening techniques to identify novel genes that induce a neural fate in AC ectodermal explants, secreted antagonists of BMPs were discovered. These proteins are secreted from the organizer (Fig. 1), and include Noggin (Zimmerman et al., 1996), Chordin (Piccolo et al., 1996), Follistatin (Fainsod et al., 1997), Cerberus (Bouwmeester et al., 1996), and XNr3 (Smith et al., 1995). They bind to BMP in the extracellular space and inhibit activation of its receptor (Fig. 3); when this occurs in the naïve embryonic ectoderm, neural tissue forms instead of epidermis (Meinhardt, 2001; Vonica and Gumbiner, 2007). Second, when expression of one of these secreted proteins in the organizer is blocked, BMP signaling is maintained on the dorsal side of the embryo and the neural plate does not form (Wessely et al., 2004). Third, when dominant-negative BMP4 protein, dominant-negative BMP receptors, or any of the BMP antagonists are expressed in AC ectodermal explants, anterior neural genes are directly induced. Similarly, when these proteins are expressed on the ventral side of embryos, a secondary axis containing neural tissue forms (Hawley et al., 1995).

Figure 3.

Pathways involved in neural induction. BMP2, 4, or 7 dimers bind to the serine-threonine receptor complex, leading to phosphorylation of the type I receptor by the type II receptor, which in turn phosphorylates (P) an appropriate Receptor regulated (R)-Smad (Smad1, 5, or 8). This phosphorylation enables the R-Smad to form a complex with the Common-Smad, Smad4, and this complex translocates to the nucleus to activate (non-neural) or repress (neural) target genes. The binding of secreted factors from the organizer, such as Noggin (Nog), to extracellular BMP prevents activation of its receptor complex. In addition, FGF and Wnt signaling lead to the inhibition of bmp4 transcription. FGF signaling also can prevent the phosphorylation of Smads1/5/8.

Mutational analysis and knock-down experiments in zebrafish, mouse, and frog embryos also demonstrate that BMP inhibition is required to form neural tissue. In zebrafish, double mutants for the BMP antagonists, chordino (a Chordin homolog) and ogon (a secreted Frizzled homolog), have a ventralized phenotype (Miller-Bertoglio et al., 1999; Yabe et al., 2003). Xenopus tropicalis embryos depleted of Chordin, Noggin, and Follistatin protein have nearly a complete loss of the central nervous system (CNS) (Khokha et al., 2005). Mouse mutants lacking the BMP antagonists, Chordin (Bachiller et al., 2000), Noggin (McMahon et al., 1998), or Cerberus (Belo et al., 2000), express only posterior neural genes, and double mutants for Noggin and Chordin fail to develop anterior brain structures (Bachiller et al., 2000). These studies indicate that across vertebrates there is a conserved and necessary role for BMP inhibition in the induction of anterior neural tissue.

There are more than 30 known BMP proteins (Balemans and Van Hul, 2002), but it is BMP4 that is mainly involved in epidermal induction and neural inhibition (Sasai et al., 1995; Wilson and Hemmati-Brivanlou, 1995; Xu et al., 1995). Its expression is restricted to the non-neural ectoderm and ventral mesoderm, and over-expression ventralizes embryos (Fainsod et al., 1994; Hemmati-Brivanlou and Thomsen, 1995). A number of BMP target genes have been identified and these include Msx1 (Foerst-Potts and Sadler, 1997; Tucker et al., 1998; Feledy et al., 1999; Ishimura et al., 2000; Takeda et al., 2000; Yamamoto et al., 2000; Yamamoto et al., 2001; Tribulo et al., 2003), Msx2 (Foerst-Potts and Sadler, 1997), Gata2 (Walmsley et al., 1994; Friedle and Knochel, 2002), Vent1 (Gawantka et al., 1995; Onichtchouk et al., 1998; Friedle et al., 1998; Rastegar et al., 1999), and Vent2 (Ladher et al., 1996; Onichtchouk et al., 1996; Onichtchouk et al., 1998; Rastegar et al., 1999; Trindade et al., 1999; Friedle and Knochel, 2002). Msx2 is not likely to be involved in inhibiting the onset of expression of early neural genes because it is not expressed until the mid-gastrula stage, which is after neural induction occurs. In contrast, Msx1, Gata2, Vent1, and Vent2 are expressed throughout the embryonic ectoderm of the animal pole of blastula embryos, and as gastrulation proceeds their expression domains are restricted to the ventral tissues in a pattern similar to that of BMP4 (Hemmati-Brivanlou and Thomsen, 1995).

Gain-of-function studies indicate that Msx1, Vent1, and Vent2 are involved in restricting the expression of neural genes to the dorsal ectoderm. Injection of Msx1 mRNA inhibited neural tissue formation induced by Noggin (Ishimura et al., 2000), and over-expression of a dominant-activator form of Msx1 prevented ventralization by BMP. However, whereas knock-down of Msx1 function with anti-sense morpholino oligonucleotides (MOs) demonstrated that it is required for the ventralizing activity of BMP4, it is not required for epidermal development, in hibition of neural induction, or axis formation (Suzuki et al., 1997; Yamamoto et al., 2000). Knock-down experiments have not been reported for the Vent proteins, but over-expression of Vent1 (Gawantka et al., 1995) or Vent2 (Ladher et al., 1996; Onichtchouk et al., 1996; Friedle and Knochel, 2002) induces epidermis and inhibits the formation of both dorsal mesoderm and neural tissue. Additionally, both Vent proteins are required to restrict expression of reporter constructs of Geminin (Taylor et al., 2006) and Sox3 (Rogers et al., 2008), two early neural ectoderm transcription factors (discussed below), to the dorsal ectoderm.

Even though inhibition of BMP signaling is sufficient to induce neural tissue in AC ectodermal explants, it is not sufficient to induce it in the ventral, non-neural ectoderm of the embryo (Wilson and Hemmati-Brivanlou, 1995; Hawley et al., 1995; Rogers et al., 2008). For example, inhibition of BMP signaling in ventral ectoderm via over-expression of Smad6, an inhibitor of the BMP effector Smad1, repressed epidermis formation but failed to induce ectopic expression of neural markers (Delaune et al., 2005; Chang and Harland, 2007). It is possible that either a second downstream pathway needs to be silenced. For example, BMP inhibition is sufficient to induce neural tissue in ventral ectoderm if signaling through both Smad1 and Smad2 are inhibited (Chang and Harland 2007). There also is evidence that additional instructive or permissive signals are required. Several experiments in frog and chick demonstrate that neural tissue is induced in non-neural ectoderm when BMP signaling is inhibited in the presence of fibroblast growth factor (FGF) (Delaune, Lemaire et al. 2005; McMahon and Moon, 1989; Christian et al., 1991a; Rodriguez-Gallardo et al., 1997; Alvarez et al., 1998; Wilson et al., 2001).

FGF signaling

In frog, there are four known FGF receptors (FGFR1-4α) implicated in neural development due to their expression in the ectoderm prior to neural induction and the subsequent restriction of their expression to neural tissue. Because FGFR2 and FGFR4α are up-regulated at the onset of neural induction, they may play a specific role in the process (Hongo et al., 1999). To determine if FGF signaling is required for neural induction, signaling was inhibited by incubation in SU5402, a chemical inhibitor of FGF signaling (Mohammadi et al., 1997), or by over-expression of a dominant-negative FGFR1 receptor (XFD) (Amaya et al., 1991) or a dominant-negative FGFR4α receptor (Δ4α) (Hongo et al., 1999). XFD inhibits FGF3 and FGF4 signaling through FGFR1, FGFR2, and FGFR3, whereas Δ4α blocks FGF8 signaling through FGFR4α (Ueno et al., 1992; Hongo et al., 1999; Bainter et al., 2001). XFD and SU5402 inhibited the induction of neural tissue in neuralized AC ectodermal explants as indicated by a loss of the pan-neural markers, Ncam and Sox2 (Launay et al., 1996; Ribisi et al., 2000; Delaune et al., 2005). Over-expression of Δ4α led to a loss of anterior neural tissue (marked by Nrp1, Bf1, and En2) in whole embryos and in neuralized AC ectodermal explants (Hongo et al., 1999). Furthermore, although dissociated AC ectodermal explants are thought to be neuralized by the reduction in levels of BMP protein due to cell dispersal, recent studies showed that BMP signaling still occurs in these cultures (Hurtado and De Robertis, 2007), and that MAP kinase signaling may drive the conversion of cells to neurons (De Robertis and Kuroda, 2004; Kuroda et al., 2005). While these studies suggest that FGF signaling is required for the formation of neural tissue in dissociated AC ectodermal explants, disruption of FGF signaling via the MAP kinase pathway by inhibiting Ras did not affect the induction of anterior neural genes, but inhibited the posteriorization of neural tissue in AC ectodermal explants (Ribisi et al., 2000). This is consistent with several previous studies indicating that FGF is required for the anterior-posterior patterning of the neural plate (reviewed in Gould and Grainger, 1997; Gamse and Sive, 2000). Moreover, treatment of embryos with SU5402 decreased some dorsal mesodermal markers, including Noggin; thus, the decrease in neural tissue may be accomplished indirectly by reducing the amount of dorsal mesoderm available to induce it (Delaune et al., 2005). Taken together, these experiments demonstrate that signaling through FGFR4α may be required for the induction of anterior neural tissue, and that FGFR1, FGFR2, and MAP kinase signaling are required for posteriorizing neural tissue.

Other studies demonstrate a role for FGF in neural induction; however, they do not demonstrate that the function of FGF is independent of its ability to inhibit BMP transcription or the BMP signaling pathway (Baker et al., 1999; Hongo et al., 1999; Ishimura et al., 2000; Wilson et al., 2000; Wilson et al., 2001; Pera et al., 2003; Sheng et al., 2003; Rentzsch et al., 2004; Wittler and Kessel, 2004; Linker and Stern, 2004; Delaune et al., 2005). In chick and zebrafish, FGF signaling inhibits BMP gene expression (Furthauer et al., 1997; Wilson et al., 2000), and in frog and zebrafish, FGF signaling inhibits BMP signaling through the phosphorylation and inactivation of its effector protein, Smad1 (Schier, 2001; Pera et al., 2003). In this manner, FGF may cooperate with or be redundant with the activity of extracellular BMP antagonists. Taken in total, these data support a role for FGF in patterning the anterior-posterior axis of neural tissues and in reinforcing the antagonism of the BMP pathway, but its requirement for neural induction independent of BMP inhibition remains in question (Lamb and Harland, 1995; Wilson et al., 2000; Takemoto et al., 2006; Rogers et al., 2008; Weisinger et al., 2008).

Wnt signaling

Maternal Wnt signaling and the inhibition of zygotic Wnt are important for the formation of neural tissue. The establishment of the dorsal-ventral axis and formation of the Nieuwkoop Center are dependent on the dorsal accumulation of the Wnt effector, β-Catenin (Larabell et al., 1997). However, it is unclear if there is a specific role for Wnt signaling in neural induction. Supporting a requirement for Wnt signaling in axis formation, over-expression of either Wnt1 or Wnt3a, which are expressed in neural tissue, induced a secondary axis (Steinbeisser et al., 1993). However, their ability to induce neural tissue in naïve ectoderm has not been tested. The strongest evidence for a role for Wnt signaling in neural induction were experiments demonstrating the induction of Ncam in AC ectodermal explants in the absence of mesoderm formation in response to mRNA injections of either Wnt8 or an activated form of β-Catenin. Both inhibited BMP4 expression, leading to the proposal that Wnt signaling sensitizes AC ectodermal explants to extracellular BMP inhibition (Baker et al., 1999). However, it is worth noting that Wnt8 normally is expressed in the ventral ectoderm, not the dorsal, presumptive neural ectoderm (Christian et al., 1991b), and in whole embryos it induces dorsal mesoderm (Otte and Moon, 1992), indicating an indirect effect. Also, loss of Wnt signaling by over-expression of a dominant-negative transcriptional effector (TCF3) or a dominant-negative receptor (Frizzled8) did not interfere with the induction of neural markers by Noggin (Rogers et al., 2008). Thus, there is little evidence for a direct role of Wnt signaling in neural induction.

In fact, studies in both frog and chick support a requirement for the inhibition of Wnt signaling for neural induction. In frog, over-expression of a dominant-active form of β-Catenin inhibited the expression of Sox2 and Sox3 in whole embryos and inhibited the induction of Sox3 by Noggin in AC ectodermal explants (Heeg-Truesdell and Labonne, 2006). Similarly, over-expression of Wnt3a in chick embryos inhibited both the neuralization of ectoderm by BMP inhibition and FGF signaling (Wilson et al., 2001). However, like FGF, Wnt signaling does play an important role in the anterior-posterior patterning of the neural plate (Gould and Grainger, 1997; Gamse and Sive, 2000).

Neural Fate Stabilization

As the presumptive neural ectoderm is established via neural inductive signaling, a large number of transcription factors is co-expressed in broad overlapping domains (Fig. 4; Sasai, 1998; Moody and Je, 2002). The mRNAs of several of these genes (Geminin, Sox3, Sox11, SoxD) are found throughout the dorsal ectoderm at the onset of gastrulation. The transcripts of others (FoxD5, Sox2, Zic1, Zic2, Zic3) are concentrated in a broad band near the blastoporal lip. The transcripts of a third group (Xiro1, Xiro2, Xiro3) are detected in two dorso-lateral bands near the blastoporal lip. Although a few are expressed maternally, most are expressed around the onset of gastrulation and continue to be expressed in the neural tube (Fig. 4). Experiments in Xenopus demonstrated that while most of these transcription factors do not induce ectopic neural tissue on their own, when they are over-expressed dorsally by mRNA injection, they all expand the neural plate. Previous studies propose that some of these genes modulate the competence of neural ectoderm to respond to signaling molecules (Sox2, Sox3, Zic1, Zic3), some maintain an immature neural state (Geminin, FoxD5, Zic2), whereas others promote the expression of neural determination/differentiation bHLH genes (SoxD, Sox11, Xiro1, Xiro2, Xiro3). How these transcription factors relate to each other to stabilize a neural fate and regulate the progression from the initially-induced neural ectoderm to neural progenitor cells has yet to be revealed.

Figure 4.

Twelve neural genes have overlapping expression patterns during the establishment of the Xenopus neural plate. (A) At the onset of gastrulation (st10), some NE genes are expressed throughout the dorsal ectoderm (orange: geminin [gem], sox3, sox11, soxD), some are expressed only in a broad dorsal band adjacent to the blastopore lip (bl) (blue: foxD5, sox2, zic1-3), and the Xiro genes (yellow) are expressed weakly in two posterior-lateral bands. Dorsal view with animal pole (AP) to the bottom. (B) The temporal expression patterns of these genes also overlap extensively. Four genes (foxD5, gem, sox3, zic2) are expressed maternally (oocyte) with detectable mRNAs through blastula stages; sox11 is expressed in oocytes but is not detected at blastula stages. The zygotic expression of most of these genes begins at late blastula (st9) to early gastrula (st10). The expression of nearly all is maintained through neural tube (nt) stages; in contrast, foxD5 is lost as the neural plate is established (st14).

Sox genes

Sox genes are named for their similarity to the testis determination factor Sry, and a conserved high-mobility group (HMG) (Sry-related HMG-box) domain that confers their DNA binding ability (Gubbay et al., 1990; Sinclair et al., 1990; Wegner, 1999). They are transcription factors that bind to the minor groove of DNA, inducing a sharp bend and regulating gene transcription (Guth and Wegner, 2008). Of the ten groups of Sox proteins, only SoxD (a Xenopus member of the SoxG group; Wegner, 1999), SoxB (Sox1, 2, 3, 14, 21), SoxC (Sox4 and 11), and SoxE (Sox8, 9, 10) groups are involved in neural induction and development (Guth and Wegner, 2008).

Sox2 and Sox3 are well documented early pan-neural markers (Uwanogho et al., 1995; Collignon et al., 1996; Penzel et al., 1997; Rex et al., 1997; Mizuseki et al., 1998a; Uchikawa et al., 1999; Graham et al., 2003; Linker and Stern, 2004; Takemoto et al., 2006; Rogers et al., 2008), and are part of the SoxB1 (Sox1, Sox2, Sox3) subfamily of transcriptional activators (Penzel et al., 1997; Uchikawa et al., 1999). They are both expressed in neural progenitor cells throughout CNS development and are required for neural progenitor maintenance (Graham et al., 2003; Ellis et al., 2004; Pevny and Placzek, 2005; Rogers et al., 2009). Their expression patterns are similar but not identical (Fig. 4). Sox3 is expressed maternally and is pan-ectodermal until the mid-gastrula stage when expression is restricted to the dorsal ectoderm (Penzel et al., 1997). In contrast, Sox2 expression begins at the onset of neural induction only in the neural ectoderm (Nitta et al., 2006). Other differences include: Sox3 is repressed in the floor plate of the neural tube prior to Sox2, expressed much more strongly in the otic placodes, and expressed in the developing lens, while Sox2 is expressed in the retina (Penzel et al., 1997; Elkins and Henry, 2006; Nitta et al., 2006; Rogers et al., 2008). These differences in their spatio-temporal expression patterns indicate that they are regulated differently, and promoter studies and bioinformatic analysis indicate that this is true (Uchikawa et al., 2003; Rogers et al., 2008). The Sox2 regulatory region has a neural induction module that is conserved across species, and in chick this module responds to both FGF and Wnt signaling for activation in the posterior neural plate (Takemoto et al., 2006). This module is present upstream of Xenopus tropicalis Sox2, but is not in the flanking sequence of Sox3. Thus far, reporter studies in transgenic frog embryos reveal that Xenopus Sox3 expression is induced by the inhibition of BMP signaling and is restricted to the neural plate by the BMP targets, Vent1 and Vent2 (Rogers et al., 2008).

Rescue experiments indicate that the SoxB1 proteins are redundant in tissues in which they are co-expressed. In mouse, loss of Sox2 in the CNS is compensated for by the up-regulation of Sox3 expression. Neurogenesis is decreased in these mice, but neural stem cells retain multipotency (Miyagi et al., 2008). In chick, over-expression of Sox1 rescues the loss of Sox2 (Graham et al., 2003). Although the SoxB1 proteins can compensate for each other, each gene is also expressed in unique tissues during development and loss in that tissue can result in severe phenotypes. Mice mutant for either Sox2 or Sox3 have relatively normal CNS formation (Malas et al., 2003; Taranova et al., 2006). However, loss of Sox2 results in diminished neurogenesis in the mouse retina (Ferri et al., 2004) and anophthalmia in humans (Fantes et al., 2003). Sox3-null mice have craniofacial abnormalities and defective pituitary development (Rizzoti et al., 2004; Rizzoti and Lovell-Badge, 2007), and Sox1-null mice suffer from epilepsy due to a complete loss of neurons in the ventral striatum (Malas et al., 2003; Ekonomou et al., 2005).

In early neural development, vertebrate SoxB1 proteins function as transcriptional activators (Bylund et al., 2003; Graham et al., 2003; Uchikawa et al., 1999). It has been proposed that they counteract neuronal differentiation induced by the SoxB2 subfamily of repressor proteins (Sox14 and Sox21) (Bylund et al., 2003). In this way, the two subgroups of SoxB proteins together may maintain a balance between proliferation and differentiation. In concordance, over-expression of both Sox2 and Sox3 in frog and zebrafish leads to a loss of neurogenesis in the cranial placodes (Dee et al., 2008; Schlosser et al., 2008) and expansion of neural progenitors in the frog and chick neural tube at the expense of epidermal development and neuronal differentiation (Kishi et al., 2000; Graham et al., 2003; Rogers et al., 2009).

The SoxC gene, Sox11, is expressed in the developing nervous system of chick (Uwanogho et al., 1995), mouse (Hargrave et al., 1997), zebrafish (de Martino et al., 2000), and frog (Hyodo-Miura et al., 2002; Brugmann et al., 2004). Sox11 is induced in response to BMP inhibition by Chordin (Hyodo-Miura et al., 2002) and by ectopic expression of the FoxD5 transcription factor (Yan et al., 2009). In Xenopus animal cap assays, Sox11 expression induces pan-neural (Ncam) and anterior neural (Otx2, En2) markers through an interaction with the MAP kinase, NLK (Hyodo-Miura et al., 2002). Sox11 is thought to be functionally redundant with the other SoxC proteins, Sox4 and Sox12, due to their overlapping expression patterns, ability to bind to the same DNA gene sequence, and weak single gene knockout phenotypes in mouse (Hoser et al., 2008; Dy et al., 2008). In chick, while Sox11 is expressed in the neural epithelium, it also is up-regulated in maturing neurons (Uwanogho et al., 1995), and functions in the differentiation of neural progenitors into neurons (Bergsland et al., 2006).

SoxD, a member of the SoxG group, is unique to amphibians (Mizuseki et al., 1998b, Guth and Wegner, 2008,). It is first detected at late blastula stages throughout the embryonic ectoderm, and like Sox3, it is then restricted to the dorsal ectoderm during gastrulation. SoxD is induced in response to the inhibition of BMP signaling, and like Sox2, can be induced in AC ectodermal explants by Zic1. Little is known about the requirement for FGF or Wnt signaling for the expression of SoxD, although the MEK5/ERK5 pathway has been implicated in its neuralizing activity (Nishimoto et al., 2005). Over-expression of SoxD in whole embryos expanded neural progenitors marked by Nrp1, induced the neural determination bHLH gene Ngnr1, and inhibited epidermal gene expression. Inhibition of SoxD by a dominant-negative construct suppressed neural tissues, a phenotype that was not rescued by Sox2, indicating that SoxD acts downstream (Mizuseki et al., 1998b).

Geminin

Geminin is a novel coiled-coil protein whose mRNA is expressed in a similar pattern as Sox3 (Fig. 4); it is maternally expressed and ubiquitous in the animal pole ectoderm until gastrulation when it is first enriched dorsally, and then restricted to the neural ectoderm (Kroll et al., 1998). After gastrulation, Geminin expression is pan-neural with an anterior bias and encompasses a region wider than that of either Sox2 or Sox3 (Kroll et al., 1998; Kroll, 2007). Geminin expression can be induced by inhibiting BMP signaling, and it is maintained in the proliferative regions of the developing nervous system (Kroll, 2007). Geminin maintains a neural progenitor population by inhibiting re-initiation of DNA replication, thereby maintaining chromosomal integrity and preventing cell cycle exit (Seo and Kroll, 2006). The latter is controlled by an antagonistic interaction with the catalytic subunits of the SWI/SNF complex, Brg1 and Brahma (McGarry and Kirschner, 1998; Seo et al., 2005). Over-expression of Geminin inhibits epidermal development and neuronal differentiation, and expands the neural progenitor population (Kroll et al., 1998; Papanayotou et al., 2008). It has been suggested that Geminin maintains cell cycling during neural plate stages and represses various aspects of differentiation, allowing for the expansion of neural ectoderm (Luo and Kessel, 2004; Pitulescu et al., 2005; Seo and Kroll, 2006).

Experiments using a human Geminin regulatory region driving GFP expression indicated that as for Sox3, Vent1 and Vent2 restrict Geminin to the dorsal side of the embryo (Kroll et al., 1998). Dorsal expression of this Geminin reporter construct required positive regulation by the Wnt signaling effector, TCF3 (Taylor et al., 2006). Additionally, recent experiments demonstrated that Geminin expression is activated directly by two other early neural proteins, Sox3 (Rogers et al., 2009) and FoxD5 (Yan et al., 2009).

FoxD5

FoxD5 is a member of the forkhead/winged helix family of transcription factors (Solter et al., 1999; Fetka et al., 2000; Sullivan et al., 2001). Although there are homologous genes in zebrafish (fkd8; Odenthal and Nusslein-Volhard, 1998), mouse (FoxD4/Fkh2; Kaestner et al., 1995), and human (FOXD5 [2q13]; Katoh and Katoh, 2004), no functional information is currently available for any of these. In Xenopus, FoxD5 maternal transcripts are localized to the animal pole, similar to those of Sox3 and Geminin. However, at gastrulation, FoxD5 zygotic transcripts are more restricted than those of Sox3 and Geminin in the presumptive neural ectoderm (Fig. 4). Unlike the other early neural transcription factor genes, FoxD5 expression is not maintained throughout neural development, but is extinguished as the neural folds elevate and fuse, except at the midbrain/hindbrain junction and the tail bud. Also, unlike the other neural genes, FoxD5 is only weakly induced by Noggin, not induced by Chordin, but is induced strongly by the maternally regulated Siamois pathway via Cerberus (Sullivan et al., 2001).

It is proposed that FoxD5, like Geminin and the SoxB1 proteins, functions to maintain a proliferating neural progenitor population. Dorsal injections of FoxD5 mRNA expanded expression of other early neural genes (Sox2, Sox3, Otx2), and repressed neural patterning (En2, Krox20) and neural differentiation (Ngnr1, NeuroD, N-tub) genes (Sullivan et al., 2001). While expression of FoxD5 induced several neural genes in AC ectodermal explants (Sullivan et al., 2001), it only induces a few of these in the ventral ectoderm of embryos (Yan et al., 2009), suggesting that it acts downstream of neural induction. Microinjections of FoxD5 dominant-activator and dominant-repressor constructs revealed that FoxD5 functions as a transcriptional repressor to expand the neural plate and deletion analysis revealed that it is dependent on the C-terminal domain of the protein. These results indicate that FoxD5 contributes to maintaining an undifferentiated neural ectoderm during the early steps of neural plate formation.

Zic genes

Three Zic genes (Zic1, Zic2, Zic3), members of the zinc finger family of transcription factors, are expressed in the early neural ectoderm (Fig. 4). All three are expressed in the presumptive neural ectoderm during early gastrulation, and later become restricted to the lateral margins of the neural plate in mouse, fish, and frog (reviewed in Moody and Je, 2002). In Xenopus, Zic1 and Zic3 are transcribed 30-60 minutes after the Chordin gene, and can be induced in AC ectodermal explants by BMP antagonism (Nakata et al., 1997; Kuo et al., 1998; Mizuseki et al., 1998a; Nakata et al., 1998), indicating that they are likely transcribed in immediate response to neural induction. Expression of Zic1 allows ectoderm to be more sensitive to neural induction by Noggin (Kuo et al., 1998), confirming its role in neural fate stabilization. Over-expression of Zic1 and Zic3 expands the neural plate and neural crest, and concomitantly represses epidermal fate. In AC ectodermal explants, Zic1 activates other early neural markers (Nrp1, SoxD) in the absence of mesoderm induction. Zic1 and Zic3 both expand the expression of neural determination/differentiation bHLH genes (Mizuseki et al., 1998a), suggesting that they promote the transition to neural differentiation.

Zic2 is expressed in a similar pattern to Zic1 and Zic3 (Brewster et al., 1998; Nakata et al., 1998), and it similarly expands the neural plate and neural crest, and reduces epidermal markers (Brewster et al., 1998). In contrast, Zic2 over-expression represses neural determination/differentiation bHLH genes, and can counteract the formation of ectopic neurons produced by Ngnr1 mRNA injection. These studies suggest that during neural plate formation Zic1 and Zic3 promote the onset of neural differentiation, whereas Zic2 maintains cells in a more immature progenitor state.

Iroquois genes

Genes within the Drosophila Iroquois complex (Iro/Irx) encode homeodomain proteins that are required for the activation of proneural bHLH genes (Gomez-Skarmeta et al., 1996; Gomez-Skarmeta and Modollel, 2002). Several vertebrate homologues have been cloned, three of which in Xenopus (Xiro1, Xiro2, Xiro3) are expressed in restricted bands in the neural ectoderm (Fig. 4) just prior to the earliest expressed neural determination bHLH genes (Ngnr1 and NeuroD). Later in development, Xiro1-3 are expressed only in the dorso-lateral neural tube and neural crest. All three Xiro genes are induced only weakly by Noggin, but their expression is greatly enhanced by the addition of posteriorizing factors, such as Wnt and FGF. These data indicate that they are expressed downstream of the afore-described early neural genes, which do not require posteriorizing factors for their induction. In Xenopus, over-expression of each of the three Xiro genes causes the neural plate to expand, and promotes the onset of neural differentiation (Bellefroid et al., 1998; Gomez-Skarmeta et al., 1998), but suppresses terminal differentiation into neurons (de la Calle-Mustienes et al., 2002). Thus, these genes are predicted to function in a position intermediate between the above-described early neural genes and the neural determination/differentiation bHLH genes (Fig. 5).

Figure 5.

A summary of the data discussed in the text predicts a gene regulatory network that results in the stabilization of neural fate. After these 12 NFS genes are induced, FoxD5 up-regulates Geminin, Sox11, and Zic2 expression. These three genes regulate each other and together (bracket) they differentially affect the expression of the other 8 NFS genes. Geminin and Sox2/Sox3 directly regulate each other, and increased Zic1 up-regulates Sox2 and SoxD. FoxD5, Geminin, Sox11, and Zic2 maintain neural ectodermal cells in an immature, stem-like state, whereas Zic1, Zic3, SoxD, and Xiro1-3 promote the onset of neural differentiation.

Do early neural transcription factors interact to regulate neural fate stabilization?

Because of their overlapping expression patterns and their common over-expression phenotype of expanding the neural plate, it seems likely that together these early neural genes coordinately stabilize the newly induced neural fate, regulate neural plate formation, and control the onset of neural differentiation. However, our understanding of how this is accomplished is woefully incomplete. Several studies indicate that one aspect of stabilizing neural fate is by modifying the signaling pathways involved in neural induction. For example: (1) in the presence of Zic1 a lower concentration of Noggin is required to induce neural ectoderm (Kuo et al., 1998); (2) Sox11 induces neural marker genes by antagonizing Wnt signaling (Hyodo-Miura et al., 2002); and (3) Geminin, Sox3, and Xiro1 each antagonize some aspect of the BMP4 signaling pathway (Kroll et al., 1998; Glavic et al., 2001; Gomez-Skarmeta et al., 2001; Rogers et al., 2008), and FoxD5 reduces the nuclear localization of the BMP signaling effectors, phosphorylated SMAD1/5/8, reduces the ventral epidermal expression of BMP4 target genes, and up-regulates the expression of Szl, a secreted BMP antagonist (Yan et al., 2009). Thus, one aspect of neural fate stabilization appears to be the regulation of the strength of signaling pathways, in particular that of BMP4, in the neural ectoderm. However, we know virtually nothing about how these several early neural genes, henceforth called neural fate stabilizing (NFS) genes, may cooperatively cause these effects.

The over-expression studies reviewed above indicate that the NFS genes can be placed into two functional groups: those that promote an immature, undifferentiated neural state (Geminin, Sox2, Sox3, FoxD5, Zic2), and those that promote the onset of neural differentiation (SoxD, Sox11, Zic1, Zic3, Xiro1, Xiro2, Xiro3). Are the genes within each group simply redundant, or do they have distinct roles in the transition from neural induction to neural differentiation? Answers to these questions are only just beginning to be experimentally addressed.

Recently, as a first step in putting the NFS genes into transcriptional order, we found that when endogenous FoxD5 levels in the neural ectoderm were reduced by targeted injection of anti-sense MOs, the expression of each of the 11 other NFS genes was down-regulated (Yan et al., 2009). This suggests that FoxD5 is necessary for either the induction or the maintenance of their expression. As expected from previous studies (Sullivan et al., 2001), over-expression of FoxD5 expanded the neural plate and thus the domains of each of the 11 other NFS genes; this was a cell non-autonomous effect because the neural plate was expanded equally well in regions devoid of the FoxD5-expressing cells, which were marked by a lineage tracer. However, when only the lineage-tagged cells expressing elevated levels of FoxD5 were examined, several responses were observed: (1) some genes that promote an immature neural fate (Geminin, Zic2) were up-regulated; (2) Sox11, which previously had been characterized as promoting neural differentiation (Bergsland et al., 2006), also was up-regulated; (3) some genes that promote an immature neural fate but are required for neural differentiation (Sox2, Sox3) were transiently and weakly repressed; and (4) genes that promote the onset of neural differentiation (Zic1, Zic3, SoxD, Xiro1, Xiro2, Xiro3) were strongly repressed. Thus, FoxD5 appears to maintain an immature neural state by differentially regulating genes that either repress or promote neural differentiation. Further experiments demonstrated that FoxD5 directly regulates Geminin, Sox11, and Zic2 transcription, and that these three genes together phenocopy FoxD5 over-expression and rescue FoxD5 MO-mediated knock-down phenotypes.

These results indicate that the NFS genes coordinately stabilize neural fate and regulate the initiation of neural differentiation, and data from several laboratories predict a NFS gene regulatory network (Fig. 5). FoxD5 acts upstream of the other NFS genes, primarily by regulating Geminin, Sox11, and Zic2. This transcriptional triad cooperatively carries out most of the effects of FoxD5 on the other NFS genes. They also regulate each other’s expression, indicating a network rather than a linear path of transcriptional regulation. There also are mutual interactions between Geminin and SoxB1 genes. Chick Geminin interacts in a complex at the N2 enhancer to promote Sox2 transcription (Papanayotou et al., 2008), and frog Sox2 and Sox3 directly regulate Geminin expression (Rogers et al., 2009). As mentioned above, Zic1 up-regulates both Sox2 and SoxD expression. In addition, those genes that promote the onset of neural differentiation (Zic1, Zic3, SoxD, Xiro1, Xiro2, Xiro3) feedback to repress FoxD5 expression, perhaps to allow the neural ectodermal cells to initiate differentiation (Yan et al., 2009). These studies are just the beginning of defining the regulatory network that promotes the transition from neural induction to differentiation. Much more work is needed to fully understand the gene interactions that comprise this network, and to determine what other factors are involved.

How Might NFS Genes Regulate Neural Stem Cells?

Stem cell therapy is expected to have widespread benefits to patients suffering from neurodegenerative diseases and neuronal loss due to trauma, stroke and congenital disease. Although the mature CNS harbors niches of self-renewing neural stem cells (Doetsch et al., 1999; Alvarez-Buylla et al., 2001), these make a significant contribution to only a few areas of brain (Cameron and McKay, 2001; Pencea et al., 2001). Therefore, many laboratories world-wide are researching how to manipulate different sources of stem cells to produce specific types of neurons for therapeutic treatments. It is most exciting that many of the NFS genes have already been implicated in controlling the differentiation of neural stem and progenitor cells. Sox2 and Sox3 are necessary for neural differentiation (Kishi et al., 2000; Wegner and Stolt, 2005; Wang et al., 2006). Each maintains neural stem/progenitor cells in a proliferative state upstream of neuronal terminal differentiation genes (Li et al., 1998; Zappone et al., 2000; Bylund et al., 2003; Graham et al., 2003; Ellis et al., 2004; Chung et al., 2006; Wang et al., 2006). Sox2 is required for adult neural stem cell maintenance and its loss causes cells to leave the ventricular zone and exit the cell cycle (Graham et al., 2003; Ellis et al., 2004; Episkopou, 2005). In contrast, Sox1 and Sox3 appear to regulate neural progenitor cell states (Kan et al., 2004; Barraud et al., 2005; Chung et al., 2006; Wang et al., 2006). Sox11 is reported to be up-regulated as neural stem cells transition to neural progenitor cells, and later it maintains pan-neural genes in neuronal progenitors downstream of bHLH differentiation factors (Uwanogho et al., 1995; Wegner and Stolt, 2005; Bergsland et al., 2006). Thus, together these Sox genes may function downstream of FoxD5, Geminin, and Zic2 to promote the initial steps from neural stem to neural progenitor cell. FoxD5 transiently represses Sox2, Sox3, and Sox11 (Yan et al., 2009), suggesting that it may act to delay the stem-to-progenitor transition. Zic genes, which promote the expression of neural determination/differentiation bHLH factors, are expressed in proliferative neural progenitor cells and are required for their expansion (Aruga et al., 2002; Ebert et al., 2003; Inoue et al., 2007). Iro/Irx genes also are involved in the initiation of expression of the neural determination/differentiation bHLH factors. Later, they play important roles in patterning the CNS (Rodriguez-Sequel et al., 2009; Stedman et al., 2009), and appear to regulate the number of cranial placode progenitors (Feijoo et al., 2009).

Conclusions

Much has been learned about the endogenous signaling pathways and transcriptional factors that induce the neural ectoderm and stabilize its neural fate. However, new experimental tactics will be needed to sort out how the various signaling pathways intersect to produce a neural fate, and likewise how the several NFS transcription factors regulate one another. Ultimately, producing neurally-committed stem and progenitor cells in sufficient quantities and at the appropriate state of differentiation for clinical applications requires that we understand the molecular mechanisms that regulate the normal process of in vivo neural development in greater detail.