By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Embryonic stem (ES) cells have been successfully used over the past decade to generate specific types of neuronal cells. In addition to its value for regenerative medicine, ES cell culture also provides versatile experimental systems for analyzing early neural development. These systems are complimentary to conventional animal models, particularly because they allow unique constructive (synthetic) approaches, for example, step-wise addition of components. Here we review the ability of ES cells to generate not only specific neuronal populations but also functional neural tissues by recapitulating microenvironments in early mammalian development. In particular, we focus on cerebellar neurogenesis from mouse ES cells, and explain the basic ideas for positional information and self-formation of polarized neuroepithelium. Basic research on developmental signals has fundamentally contributed to substantial progress in stem cell technology. We also discuss how in vitro model systems using ES cells can shed new light on the mechanistic understanding of organogenesis, taking an example of recent progress in self-organizing histogenesis.
Embryonic stem cells as powerful tools for developmental biology
Pluripotency is generally defined as the cellular ability to differentiate into three germ-layer derivatives (Niwa 2007). Inner cell mass (ICM) and epiblast of the mammalian early embryo are typical pluripotent tissues. They are, in a sense, equivalent to the animal cap of the amphibian embryo, which has been widely used in classical and modern embryology. However, unlike the amphibian animal cap, the ICM and epiblast of mammalian embryos are difficult tissues to handle and prepare in large amounts.
Embryonic stem cells, established from the ICM of blastocyst, can differentiate into all kinds of embryonic tissues, mimicking the pluripotent nature of the origin. Over the last decade, numerous studies have demonstrated steered differentiation of embryonic stem (ES) cells into various tissues by mimicking the signaling environments of the early embryo (Wichterle et al. 2002; Mizuseki et al. 2003; Gotz & Barde 2005; Murry & Keller 2008; Yeo et al. 2008; Zhang et al. 2008). A straightforward application of these techniques is to use these differentiated cells for regenerative medicine and drug discovery. Another equally important application is to use these culture systems in embryogenesis and organogenesis studies. Recent progress in molecular biology has identified genes and molecules responsible for particular developmental processes. In this direction of study, a standard approach is to analyze the defects appearing after deletion of a gene or molecule from the whole system, which is effective for identification of an essential element. However, the elucidation of complex interactions of multiple components, as seen in organogenesis, is not simply achievable in such an approach.
The in vitro differentiation culture of ES cells may provide a complementary system involving constructive approaches. For instance, the system is suitable for finding minimum requirements for recapitulating the differentiation of target tissues in vitro. Since ES cells are homogenous and do not carry much predisposition, unlike the amphibian animal cap that includes maternal information, the effects of exogenous patterning signals can be more precisely analyzed. The use of ES cells is particularly advantageous for analyzing the regulatory mechanisms in early mammalian embryogenesis, which is technically demanding to study in classical embryological experiments.
In the following sections, we discuss recent progress in neural differentiation of ES cells and its contribution to progress in developmental neurobiology.
Default model: Neural induction and early regional patterning
Neural induction represents the earliest step in the determination of the neural fate in the ectoderm. Spemann and Mangold (1924) demonstrated that the dorsal lip of the amphibian blastopore (Spemann’s organizer), which gives rise mainly to the axial mesoderm, emanates inductive factors that direct neural differentiation from the uncommitted ectoderm. The molecular nature of these neural inducers had been veiled in secrecy for more than six decades after this first study. In the mid-1990s, several molecules with direct neural activity were identified, including Noggin, Follistatin and Chordin (Smith & Harland 1992; Lamb et al. 1993; Hemmati-Brivanlou & Melton 1994; Sasai et al. 1994, 1995). These neural inducers turned out to be bone morphogenetic protein (BMP)-binding molecules that antagonize BMP signaling (Piccolo et al. 1996; Zimmerman et al. 1996; Fainsod et al. 1997). BMP4 strongly inhibits neural induction and can restore epidermal fates in dissociated Xenopus ectodermal explants (Wilson & Hemmati-Brivanlou 1995). Independently, several studies demonstrated that dissociated naïve ectodermal explants from Xenopus blastula-stage embryos, which normally give rise to epidermis, would become neural tissue in the absence of “inducing” signals from the organizer (Godsave & Slack 1989; Grunz & Tacke 1989; Sato & Sargent 1989). These findings led to the “default model” of neural induction (Hemmati-Brivanlou & Melton 1997; Sasai & De Robertis 1997), proposing that ectodermal cells have an autonomous tendency to differentiate into neural progenitor, unless their differentiation is inhibited by BMP signaling. FGF signaling is also implicated in neural induction (particularly in chick development), although their precise contribution has remained somewhat controversial (Stern 2005, 2006).
Understanding of mammalian neural induction was not substantially advanced, as compared with that of amphibian induction, mainly because of the technical difficulty of analysis. A strong support for the validity of the neural default model in mammalian neural development came from ES cell studies. Mouse ES cells selectively differentiate into neural progenitors when they are cultured in serum-free and patterning factor-free conditions (Tropepe et al. 2001; Ying et al. 2003; Watanabe et al. 2005; Smukler et al. 2006). Based on this idea, for instance, we established an efficient culture system for selective neural differentiation of mouse and human ES cells, namely, serum-free culture of embryoid body-like aggregates with quick reaggregation (SFEBq; Wataya et al. 2008; Eiraku et al. 2008). In this culture system, ES cells are dissociated (to minimize possible effects of the culture substrate or matrix) and subsequently reaggregated using a low-cell-adhesion 96 well plate. Neural differentiation in the SFEBq culture spontaneously occurs and does not require extrinsic neural inducers. ES cells selectively differentiate into neural progenitors (>95% in total cells). Wnt inhibitors (e.g., dickkopf-1) and Nodal inhibitors (e.g., lefty-1) may be added to minimize endogenous signals. Inhibition of endogenous SMAD signaling with transforming growth factor (TGF)-β antagonists can promote neural induction, particularly in human ES and induced pluripotent stem (iPS) cells (Chambers et al. 2009). These results are consistent with the “neural default model” advocated in amphibian development. Recent findings suggest that the zinc-finger protein Zfp521 drives the initial neural specification from ES cells by activating genes associated with early neural development (Kamiya et al. 2011).
An important feature of ES cell-derived neural progenitors is that they adopt the rostral forebrain fate unless they receive caudalizing signals (Wichterle et al. 2002; Li et al. 2005; Watanabe et al. 2005; Elkabetz et al. 2008), in accordance with the “Nieuwkoop’s activation-transformation model” (Nieuwkoop 1952; Stern 2001; Rallu et al. 2002; Wilson & Houart 2004; Levine & Brivanlou 2007). For instance, SFEBq-cultured mouse and human ES cells acquire a rostral forebrain fate (Sox1+/Six3+), and ES cells efficiently generate telencephalic progenitors (Foxg1+) in the presence of Wnt inhibitors (Watanabe et al. 2005, 2007; Eiraku et al. 2008). A similar result is also observed when ES cells are cultured in medium devoid of serum or any morphogenetic factors (Gaspard et al. 2008). Interestingly, when ES cells are cultured in completely growth factor-free medium, not containing even insulin, ES cells differentiate into rostral hypothalamus (Wataya et al. 2008), consistent with the rostral hypothalamus may be assigned as the rostral-most region of the neural tube (the exact assignment of the AP axis in the brain is still under debate though). Given that the neuronal repertoire of the rostral hypothalamus is highly conserved in the evolution of the central nervous system, even in vertebrates and invertebrates (Tessmar-Raible et al. 2007), the rostral hypothalamic region might be relevant to the default status of the naïve neuroectoderm in vivo (Wataya et al. 2008).
Regional specification along the antero-posterior and dorso-ventral neural axes
During development, the nervous system generates a variety of neurons that have different functions, depending on their position along the anterior-posterior (AP) and dorsal-ventral (DV) axes. In other words, the cell fate of the neural progenitor is largely dependent on the positional information provided by extracellular cues (e.g., secreted signaling molecules). Along the AP axis, the neural tube is subdivided into the prosencephalon (forebrain), the mesencephalon (midbrain), the rhombencephalon (hindbrain) and the spinal cord (Lumsden & Krumlauf 1996). The early neuroectoderm of the neural tube is patterned along the AP axis by the influences of various caudalizing signals and their antagonists, while the signals’ effects differ depending on timing, combination and concentration (Nieuwkoop 1952; Muhr et al. 1999; Stern 2001). Some of these factors are produced from the neighboring mesoderm (notochord, prechordal plate and paraxial mesoderm) and endoderm (anterior visceral endoderm) (Echevarria et al. 2003; Vieira et al. 2010). In addition, two well-known signaling centers for early AP patterning exist within the neural tube (Wurst & Bally-Cuif 2001). One is the anterior neural ridge (ANR), which is located at the junctional region between the prosencephalon and the anterior non-neural ectoderm and is required for the maintenance of the early forebrain identity. The ANR produces Fgf8, which promotes the expression of Foxg1 (Bf1), a forebrain-specific transcription factor (Shimamura et al. 1997). Chordin and Noggin are also expressed in the ANR and support Fgf8 expression by antagonizing local BMP signals (Anderson et al. 2002). The other is the isthmic organizer, which lies at the midbrain-hindbrain boundary (MHB). The isthmic organizer secretes patterning signals such as Wnt1 and Fgf8 and controls the AP pattern of the mesencephalon. In addition, the isthmic organizer is essential for the induction of cerebellar development at the rostral end of the hindbrain, while its transplantation in the rostral midbrain can cause ectopic formation of cerebellar structures from the neighboring tissue. The isthmic organizer is formed and maintained by the intricate regulatory functions of region-specific transcription factors such as En1/2, Pax2/5/8, Otx2 and Gbx2. The expression of these factors is under the control of Fgf8 secreted by the isthmic organizer itself (Wurst & Bally-Cuif 2001), and the format of a positive feedback loop involves Fgf8 contributing to the maintenance of the tissue’s identity.
In addition to Fgf8, retinoic acid (RA) and Wnt also contribute to the AP patterning of neural tube. Both RA and Wnt are derived from the paraxial mesoderm with a gradient posterior-high/anterior-low (Maden 2002; Nordsrtöm et al. 2002). In the AP axis of the neural tube, RA and Wnt, along with Fgf, are responsible for the specification of the midbrain, the hindbrain and the rostral spinal cord (the fate specification of the caudal spinal cord still remains elusive).
In amniotes neural development, the regional pattern along the DV axis becomes fixed subsequently to the AP patterning (i.e., soon after neural-tube closure). The basic mechanism of DV patterning seems to be conserved almost throughout the neural tube. Except in the rostral forebrain, which is highly complex in structure, dorsalization is promoted by BMP and Wnt signaling, which emanate from the epidermis and the roof plate, while ventralization is caused by Shh, which is produced from the notochord and the floor plate (Echelard et al. 1993; Lee & Jessell 1999; Hoch et al. 2009; Hansen et al. 2011) (Fig. 1).
Ample studies have demonstrated that ES cell-derived neural progenitors have the ability to perceive positional information of patterning signals. As described above, neural progenitors derived from ES cells intrinsically acquire the rostral forebrain identity, while caudal fates, such as those of midbrain, hindbrain and spinal cord, can be induced by treating with Fgfs, RA and Wnts (Lee et al. 2000; Wichterle et al. 2002; Mizuseki et al. 2003; Watanabe et al. 2005; Bouhon et al. 2006; Su et al. 2006; Salero & Hatten 2007; Gaspard & Vanderhaeghen 2009; Germain et al. 2010). In cell culture, other signaling molecules can also have some effects on AP patterning. An intriguing finding in this line is that insulin, a commonly used additive in various culture media, has moderate but substantial caudalizing effects on ES cell-derived neural progenitors. Indeed, the addition of insulin suppresses the expression of Six3 and Rx (rostral forebrain markers), and increases the expression Fgf8 and En2 (expressed in the MHB region) (Wataya et al. 2008; Muguruma et al. 2010), although the exact in vivo role of insulin (or IGF) signaling in mammalian neural patterning remains to be elucidated.
Cerebellar neurogenesis from mouse ES cells: An example of positional control
The cerebellar cortex is a highly ordered structure composed of three distinct layers; molecular layer, Purkinje cell layer, and granule cell layer. During embryonic development, by the inductive influence of the MHB, the cerebellum anlage arises from two distinct tissues in the dorsal region (alar plate) of rhombomere 1: the cerebellar ventricular zone, which expresses the basic helix-loop-helix (bHLH) transcription factor Ptf1a, and the rhombomere 1 rhombic lip (upper rhombic lip), which expresses another bHLH factor, Atho1 (Math1). The Ptf1a+ progenitors produce GABAergic neurons of the cerebellar cortex (Purkinje, basket, stellate and Golgi neurons) and of the deep cerebellar nuclei (DCN). In contrast, the Atho1+ progenitors generate cerebellar glutamatergic neurons, including granule cells and large DCN projection neurons (Hoshino 2006) (Fig. 2). The initial phase of cerebellar development depends on the formation and function of the isthmic organizer (Zervas et al. 2005). Fgf8 is an important signaling molecule produced by the isthmic organizer. Loss-of-function and gain-of-function of Fgf8 induce severe cerebellar defects and ectopic cerebellar formation, respectively. However, treatment of ES cell culture with Fgf8 is not sufficient for efficient induction of cerebellar tissue.
To effectively generate cerebellar tissue from ES cells in SFEBq culture, we took the following strategy for mimicking the endogenous program for cerebellar development by the isthmic organizer. First, we searched for conditions to caudalize the ES cell-derived neuroectodermal progenitors in SFEBq culture for generating the isthmic-organizer tissues and found that the treatment with insulin and Fgf2, which have weak caudalizing activity, at an early culture phase (days 1–2) is suitable for this purpose. Treatment with insulin and Fgf2 synergistically promoted Wnt1 and Fgf8 expression and effectively induced En2+ tissues (MHB and cerebellar tissues; >80% of total cells). In this case, we do not think that the combination of insulin and Fgf2 is very specific for the induction of MHB and cerebellar tissues; instead, this culture condition is probably permissive for a broad mid-hindbrain regionalization in ES cell-derived neuroectoderm and does not interfere with the autoregulatory positive feedback loop involving Wnt1 and Fgf8. Therefore, once a small amount of isthmic organizer tissue is formed in the ES cell aggregate, the organizer tissue could grow into a substantial mass that is sufficient to induce cerebellar tissues in the rest of the aggregate.
The next step is to give an appropriate DV identity to the neural progenitor. The DV positional information can be modified by treating the culture with dorsalizing factors (e.g., BMP4, Wnts) and ventralizing factors (e.g., Shh), or with their antagonists. In SFEBq culture system, the dorsal specification of En2+ tissues into cerebellar-plate progenitors could be achieved by using the hedgehog antagonist from days 7–9. This is a passive dorsal specification in the sense that the ventralizing signal (Shh) is blocked, instead of giving an active dorsal specification. This moderate dorsalization is just enough to cause a marked generation of Purkinje cell precursors expressing the transcriptional co-repressor Corl2+ (40% of total cells). In contrast, treatment with BMP4, a strong dorsalizing factor, may be too intense. It induced the differentiation of rhombic lip tissue, which generates Atho1+ granule cells and Tbr1+ DCN neurons (both are glutamatergic neurons), while the induction of Purkinje cell precursors (alar-plate derivatives) was not efficient in this treatment. Collectively, the timed treatment with insulin + Fgf2 in SFEBq culture facilitates ES cells to acquire the neural fate of the rhombomere 1-MHB identity by activating the endogenous isthmic organizer program and the subsequent hedgehog inhibition promoted cerebellar-plate specification. This is a successful example of steering of ES cells by a combinatory control of AP/DV-axis value and intrinsic signaling networks.
Histogenesis of neural tissues by self-organizing activity
Once the regional fate is specified within the neuroectoderm by global positional information, local programs start functioning to generate a variety of region-specific characteristics. Neurogenesis is precisely regulated both spatially and temporally. Specific types of neurons are generated at the right time and in the right position. Newly generated neurons, leave the ventricular zone for their final position, acquire polarization (e.g., axon vs. dendrite) and form nuclei or layers in the brain. In the embryo, neural tissues form highly ordered structures that exhibit region-specific characteristics. Can the ES cell culture system be used for recapitulating the local neural tissue formation in vitro?
Recent studies indicate an unexpected level of intrinsic self-organizing activity of ES cell-derived neural tissues. One of the remarkable features of the SFEBq culture is the reproducible formation of the well-polarized neuroepithelial structures within the ES cell aggregates (Eiraku et al. 2008, 2011; Muguruma et al. 2010). N-cadherin+ neural progenitors formed a characteristic neuroepithelial structure. Although the neuroepithelium tends to break into smaller spheres (neural rosettes) in longer culture, each neural rosette has a typical apico-basal polarity of neuroepithelium. The apical side markers (gamma-tubulin, CD133/prominin, aPKC) are localized on the inner side facing the central lumen of the rosette, where cells under mitosis (detected by phosphorylated histone H3 are found). The basement membrane (laminin+) forms on the basal (outer) surface. This topology is similar to that of the embryonic neuroepithelial structure (Fig. 3).
For the mechanism of generating diverse neurons in a stereotypical fashion, sequential neurogenesis and gene expression according to spatiotemporal change are the principle concepts, particularly in layered neural tissues such as cerebral cortex, retina, and cerebellum. In fact, the polarized neuroepithelial tissues formed in SFEBq culture can recapitulate the organization of the in vivo tissues. First, we would like to take an example of cerebellar cortex for histogenesis study (Muguruma et al. 2010). Multiple types of neurons with a unique morphology are born in two spatially separated germinal zones (the ventricular zone and the rhombic lip) and migrate into the cerebellar primordium, called the cerebellar plate. At this stage, different types of neurons appear to be intermingled. Then, they are gradually sorted and finally separated into discrete layers. In the early cerebellar primordium (day E11–12), the neuroepithelium begins to express tissue-specific marker genes, Ptf1a and Neph3 (Kirrel2; a membrane protein). Ptf1a and Neph3 are expressed together in most of the cerebellar plate neuroepithelium. In the cerebellar plate neuroepithelium, Purkinje cells and cerebellar GABAergic interneurons are born at distinct developmental stages: E11–13 for Purkinje cells and E13–P15 for GABAergic interneurons (Pax2+) in mice (Altman & Bayer 1997; Maricich & Herrup 1999). Postmitotic Purkinje cell precursors become positive for Corl2 in the subventricular zone. The portion of the cerebellar plate neuroepithelium generating Purkinje cells co-expresses E-cadherin during E12–14 (Figs 2, 3).
The spatial expression of differentiation markers of Purkinje cells in SFEBq culture resembles the in vivo expression pattern. On day 13 of SFEBq culture, N-cadherin+ neural cells form epithelial rosettes with a clear apical-basal polarity and are proliferative (Ki67+). In these neural rosettes, a large majority of the cells were double-positive for Neph3 and Ptf1a, while cells under mitosis (pH3+) are located at the apical end as in vivo. Corl2+ Purkinje cell precursors (post-mitotic) are found basal to (just outside of) each epithelial rosette, surrounding the Ptf1a+ proliferative zone.
SFEBq-cultured cells recapitulate not only spatial expression of neural markers but also temporal aspects of the in vivo development of cerebellar neurogenesis. In the embryonic cerebellar plate, the subtype specification of GABAergic neurons (Corl2+ Purkinje cells and Pax2+ interneurons) is strongly linked to the birth date of neurons: early-born neurons differentiate into Purkinje cells (E11–13), whereas late-born neurons become GABAergic interneurons (E13–P15). In SFEBq culture, birth-dating analysis using EdU showed that Corl2+ cells are born more frequently on day 13 than on day 15. In contrast, Pax2+ cells were more efficiently labeled by EdU on day 15 than on day 13, suggesting that Purkinje cell generation is regulated in a similar manner to that in the embryo. When day 13 Neph3+ cerebellar progenitors are isolated from culture by fluorescence-activated cell sorting (FACS), these cells differentiate efficiently into Corl2+ neurons in a few days and eventually into fully arborized large Purkinje cells that express specific markers such as L7, calbindin, cerebellin and GluRdelta2, and exhibit characteristic electrophysiological properties. Thus, precise temporal regulation of the SFEBq culture system is essential for cerebellar neurogenesis in vitro.
Recapitulation of neurogenesis is not sufficient for histogenesis in vitro. Newly generated neurons must be added and integrated into the pre-existing tissues as components of the system. Such ability can be tested by transplantation of ES cell-derived neurons into the neural tissue of the animals. When day 13 Neph3+ cerebellar progenitors are injected into the subventricular space of the E15.5 mouse cerebellar plate, the majority of grafted neurons surviving in the host cerebellar plate express Purkinje cell-specific markers (L7, calbindin) 4 weeks after transplantation. These integrated Purkinje cells show the normal cell polarity, and have well-developed dendrites growing outward in the molecular layer (Fig. 4a). Their primary dendrites are surrounded by multiple VGluT2+ spots (Fig. 4b), which represent presynaptic inputs from the climbing fibers. In addition, the location of GluRdelta2 in the periphery of the dendrites is juxtaposed to that of cerebellin (Cbln1), a marker for parallel fibers, indicating that grafted cells receive parallel fiber inputs from the host cerebellum (Fig. 4c). The grafted Purkinje cells have a long axon extending through the granule layer and the white matter toward the DCN (Fig. 4d). These observations suggest that ES cell-derived Purkinje cells can be integrated into the host tissue and join the formation of functional neural circuits by receiving inputs and sending outputs.
The ability of tissue integration depends both on the properties of the grafted cells and on the environment of the host tissue, while dynamically changes during ontogenesis. When Neph3+ progenitors were isolated on day 15, instead of day 13, they mostly became GABAergic interneurons rather than Purkinje cells. When postnatal mice were used as a host of transplantation, the integration efficacy is substantially low as compared with the fetal graft. Thus, integration of Purkinje cells requires strict temporal matching of the grafted cells and the host tissues.
The self-developing system in SFEBq culture can be used to generate the organized tissue seen in early neurogenesis of specific brain regions such as cerebellum (Muguruma et al. 2010). A similar self-organizing formation of polarized neuroepithelium is observed in SFEBq culture for the generation of cerebral cortex tissue (Eiraku et al. 2008). In this case, mouse and human ES cells are coaxed to differentiate into cerebral cortex tissue in low-growth-factor medium. Interestingly, the cortical progenitors in this culture formed stratified neuroepithelium with a ventricular, subventricular, cortical plate and Cajal-Retzius-cell zones. However, unlike the postnatal neocortex, the formed layers fail to generate the inside-out pattern (in which late-born neurons migrate to more superficial layers than early-born neurons), indicating that some key components are still missing in the culture.
The formation of a more complete stratified structure was recently reported for neural retina development in SFEBq culture (Eiraku et al. 2011). In this case, retinal differentiation was induced in low-growth-factor medium supplemented with matrigel, and the retinal neuroepithelium spontaneously formed an optic-cup structure in a self-organizing manner. The ES cell-derived neural retina tissue developed into a fully stratified architecture (photoreceptors, horizontal cell, bipolar cell, amacrine cell, ganglion cell and Müller glia) as seen in the postnatal eye, demonstrating the high potential of the intrinsic self-developing program. Another recent report demonstrated that SFEBq culture could be used to generate functional anterior pituitary tissue producing pituitary hormones, which rescued the phenotypes of hypopituitary mice (Suga et al. 2011). The three-dimensional culture system of ES cells may open new avenues in developmental biology and regenerative medicine.
Embryonic stem cell systems have been considered as potential tools to generate specific types of neurons in vitro. Knowledge and technology brought by ES cell research provide technical bases for regenerative medicine and applied medical research such as pathogenesis and drug discovery. The recapitulation of highly ordered embryonic tissues in three-dimensional ES cell culture systems will be powerful tools also for basic mechanistic studies in developmental neurobiology.
This work was supported by grants-in-aid from the Ministry of Education, Culture, Sports Science and Technology of Japan (K.M.) and the National Center of Neurology and Psychiatry (K.M.).