Address correspondence and reprint requests to Joshua G. Corbin, PhD and Tarik F. Haydar, PhD, Center for Neuroscience Research, Children’s National Medical Center, Washington, DC 20010, USA. E-mail: firstname.lastname@example.org and email@example.com
The mammalian telencephalon, which comprises the cerebral cortex, olfactory bulb, hippocampus, basal ganglia, and amygdala, is the most complex and intricate region of the CNS. It is the seat of all higher brain functions including the storage and retrieval of memories, the integration and processing of sensory and motor information, and the regulation of emotion and drive states. In higher mammals such as humans, the telencephalon also governs our creative impulses, ability to make rational decisions, and plan for the future. Despite its massive complexity, exciting work from a number of groups has begun to unravel the developmental mechanisms for the generation of the diverse neural cell types that form the circuitry of the mature telencephalon. Here, we review our current understanding of four aspects of neural development. We first begin by providing a general overview of the broad developmental mechanisms underlying the generation of neuronal and glial cell diversity in the telencephalon during embryonic development. We then focus on development of the cerebral cortex, the most complex and evolved region of the brain. We review the current state of understanding of progenitor cell diversity within the cortical ventricular zone and then describe how lateral signaling via the Notch-Delta pathway generates specific aspects of neural cell diversity in cortical progenitor pools. Finally, we review the signaling mechanisms required for development, and response to injury, of a specialized group of cortical stem cells, the radial glia, which act both as precursors and as migratory scaffolds for newly generated neurons.
The remarkably complex mammalian CNS, with its diverse cell types and billions of connections, arises from the neural plate, a simple group of cells in the early embryo. This single layer of neural stem cells gives rise to a multitude of neuronal and glial cell types that will populate the mature mammalian CNS. How this process occurs, from the early patterning to the specification of distinct neuronal and glial cell types, has become much clearer over recent years. A major emerging theme is that different compartments within the CNS share similar developmental strategies. Much of our understanding of the generation of neural cell diversity in the telencephalon has been based on concepts that materialized from studies of spinal cord development in both chick and mouse. One finding with broad implications for all of CNS development is that progenitor pools generating different spinal cord neuronal and glial cell types are initially specified in response to gradients of early expressed morphogens (around the time of neural tube closure), most notably, sonic hedgehog (Briscoe et al. 2000; Placzek and Briscoe 2005). The read-out of morphogen gradients is the expression of combinations of transcription factors (patterning genes) along the dorsal/ventral axis in the spinal cord ventricular zone (VZ). The combinatorial expression of these transcription factors, which are typically of the homeodomain and basic helix-loop-helix (bHLH) classes, instructs each unique progenitor population (or progenitor pool) to generate progeny that are committed to specific neural fates (Zhou and Anderson 2002; Sugimori et al. 2007). This ‘combinatorial code’ also allows large amount of progenitor cell diversity to be generated from a relatively small number of genes (e.g. a combinatorial code that is based on four genes theoretically allows for 16 different combinations of cell types to be generated). Both the dorsal/ventral location (distance from source of secreted patterning factors) and the temporal expression of these genes are key determinants for the competency of progenitor cells to generate different cell types over time. For example, during early time points of mouse embryogenesis [approximately embryonic day (E) 9.5], the pMN spinal cord progenitor domain in the ventral spinal cord first gives rise to motoneurons. Later, at about E12.5, this progenitor pool gives rise to oligodendrocytes. This switch appears to occur via the differential temporal expression of the bHLH genes Olig2 and Neurogenin1/2 (Kessaris et al. 2001; Rowitch 2004).
Specification of neural cell diversity in the telencephalon appears to follow the same basic developmental logic as in the spinal cord. First, the early patterning mechanisms appear to be conserved. For example, sonic hedgehog, along with other secreted factors such as Wnts, fibroblast growth factors, and bone morphogenetic protein (BMPs), play an essential role in the parcellation of the early telencephalic VZ into separate progenitor pools (Fuccillo et al. 2006; Aboitiz and Montiel 2007). Second, in ways that are not yet completely understood, these signaling molecules and growth factors also regulate the expression of unique combinations of transcription factors in these progenitor cells. Interestingly, many of these are the same transcription factors (e.g. Pax6, Gsh2) or are of the same class (e.g. Nkx family) that perform such instructive roles in the spinal cord.
Nevertheless, despite these similarities, there are notable added complexities to telencephalic development that reflect the richness of cell diversity in the telencephalon when compared with the spinal cord. For example, telencephalic inhibitory GABAergic neurons are highly diverse, and based on a combination of morphological, immunohistochemical and electrophysiological criteria, can be subdivided into at least 20 different subtypes (Parra et al. 1998). This diversity allows for the formation of highly sophisticated circuitry that can vary from telencephalic structure to structure. Over the past 10 years, a series of dye-labeling, gene knockout, and cell transplantation studies have revealed that the vast majority (if not all) of GABAergic interneurons (and GABAergic projection neurons) are generated in the ganglionic eminences of the ventral (subpallial) telencephalon (Fig. 1). This is in contrast to excitatory, glutamate projection neurons which are generated in the germinal zones of the cerebral cortex and hippocampus in the dorsal (pallial) telencephalon (Corbin et al. 2001; Marin and Rubenstein 2003; Wonders and Anderson 2006). Recent genetic loss-of-function and fate mapping studies have begun to unravel the details of the spatial and temporal generation of this interneuronal diversity (the extent of an embryonic contribution to functional excitatory neuronal diversity remains unexplored). This work reveals that different interneuronal cell types are generated within spatially separate progenitor cell compartments of the ganglionic eminences, and also at different times during embryogenesis (Fig. 1). Specifically, the medial ganglionic eminence (MGE) is the major source of interneurons, giving rise to the morphologically-heterogeneous non-fast spiking somatostatin-positive interneurons and fast spiking parvalbumin-positive (PV+) subsets of cortical interneurons (Wichterle et al. 2001; Xu et al. 2004; Butt et al. 2005; Fogarty et al. 2007). In contrast to the MGE, the caudal ganglionic eminence (CGE) appears to provide a more restricted subset of interneurons, giving rise only to the calretinin-positive (CR+) bipolar interneurons and double bouquet cells (Fig. 1) (Nery et al. 2002; Xu et al. 2004; Butt et al. 2005).
Recent studies have also revealed the MGE and the CGE can be further parcellated into smaller subdomains based on combinatorial expression of both homeodomain and bHLH transcription factors (Flames et al. 2007). Indeed, it has recently been shown that the dorsal and ventral components of the MGE, which express unique combinations of transcription factors, differentially give rise to somatostatin-positive and PV+ neurons respectively (Flames et al. 2007; Wonders et al. 2008). The lateral ganglionic eminence (LGE) can also be subdivided into at least two distinct dorsal LGE and ventral LGE progenitor pools (Stenman et al. 2003). The ventral LGE appears to be a major source of striatal GABA projection neurons while the dorsal LGE is a source of olfactory bulb calretinin (CR+) neurons as well as subclass(es) of neurons in the piriform cortex and amygdala (Carney et al. 2006; Waclaw et al. 2006; Garcia-Moreno et al. 2007). Recent studies employing the above techniques have also revealed that interneuronal subtype identity is further refined over time. For example, while the CGE generates broadly similar classes of CR+ interneurons at both E13.5 and E15.5, there are distinct differences in the specific electrophysiological firing properties of cells generated at these time points (Miyoshi et al. 2007). Thus, in a mechanism highly reminiscent of the spinal cord, neural cell diversity in the telencephalon is generated by a combination of progenitor pool location and changes in gene expression over time.
To generate regional cell diversity in the mature telencephalic structures these neural cells are exported from their birth places. With regard to cerebral cortical development, the most well studied region of the telencephalon, interneurons migrate tangentially from the subpallium into the dorsal neocortical wall where they meet the radially migrating excitatory neurons which are generated locally in the dorsal VZ. In contrast, structures located more ventrally in the telencephalon, such as the amygdala, import the dorsally generated excitatory neurons via a route of migration known as the lateral cortical stream (Carney et al. 2006). Thus, the embryonic telencephalon is characterized by large populations of cells moving in a variety of complex and sometimes convoluted directions (Fig. 2).
The vast majority of these cells are post-mitotic as they migrate, indicating that they are already fated, even to their specific subtype, before embarking on their journey. This ‘pre-determination’ is supported by a number of lines of evidence. The first is heterotopic transplantation studies in which labeled MGE or CGE progenitor cells introduced into ectopic locations retain both their migratory routes and ultimate positional and specific interneuronal subtype fate in the mature brain (Wichterle et al. 2001; Nery et al. 2002; Butt et al. 2005). For example, when ectopically transplanted into the MGE, CGE precursors maintain their fate and differentiate into CGE-derived interneuron subtypes (CR+ or vasointestinal peptide-expressing bipolar and double bouquet interneurons). Second, when plated in culture on early post-natal cortical feeder layers or transplanted directly into the early post-natal cortex, ganglionic eminence precursors also maintain their ability to differentiate into their pre-determined phenotypes (Xu et al. 2004; Alvarez-Dolado et al. 2006). Third, in vitro work has revealed that cortical PV+ interneurons differentiate and form appropriate stereotypical synapses onto excitatory cells in the absence of thalamic input (Di Cristo et al. 2004). Thus, it appears that major aspects of inhibitory neuronal subtype fate and connectivity appear to be specified locally within the progenitor zones. However, the extent to which local cues act upon these immature neurons to guide their full differentiation into their respective mature cell types and sculpt their connectivity remains unknown.
In addition to generating the post-mitotic neurons of the mature brain, subsets of these progenitor pools also seed adult progenitor cell niches. In the adult brain, there are two known persistent sources of progenitor cells; the subventricular zone (SVZ) of the lateral ventricle and the subgranular zone of the hippocampal dentate gyrus (Alvarez-Buylla and Lim 2004; Ninkovic and Gotz 2007). The SVZ begins to generate olfactory bulb interneurons around mid-neurogenesis and continues to make both these interneurons as well as oligodendrocytes through adulthood (Belachew et al. 2001; Marshall et al. 2003). In the adult brain, the SVZ is a complex mix of stem and progenitor cells that express a variety of both pallial and subpallial markers such as Gli1, Pax6, Emx1, and Dlx2. Consistent with the differential expression patterns of these markers found during embryonic development, genetic fate mapping studies reveal that the unique stem cell niche in the post-natal SVZ is in fact derived from early embryonic progenitor pools from both the pallium and subpallium (Ahn and Joyner 2005; Kohwi et al. 2007; Young et al. 2007). Thus, events regulating the generation of neural cell diversity during embryonic development also appear to be involved in the generation and maintenance of progenitor cells in the adult brain. The recent advances in understanding prenatal precursor dynamics as further described below may also shed light on post-natal progenitor cell biology.
The generation of embryonic cerebral cortical precursor cell diversity
Astoundingly, the diverse multitude of cells in the mammalian telencephalon arises during embryonic development from a single layer of neuroepithelial stem cells which form the VZ. As such, the cellular constituency of the VZ and how these cells generate such diversity have long been central questions of neuroscience. Indeed, the debate over the precise morphology and lineage potential of VZ cells began almost a century ago and continues today.
In the dorsal (pallial) telencephalon, the VZ is a pseudostratified neuroepithelium initially composed entirely of neuroepithelial stem cells that divide rapidly and symmetrically to expand the population of neocortical founder cells. At the onset of neurogenesis, neuroepithelial cells are thought, at least in part, to transition into a specialized VZ cell type called radial glial cells (RGCs), which begin dividing asymmetrically to produce self-renewing RGCs and post-mitotic neuronal daughter cells with each round of division (Gotz and Huttner 2005). While RGC somata are contained within the VZ, their long radial processes span the neocortical wall and terminate in endfeet at both the ventricular and pial surfaces (Fig. 3). Importantly, these long processes remain attached to these surfaces even during mitosis (Rakic 1972; Miyata et al. 2001; Noctor et al. 2001). The maintenance of RGC morphology is critical for proper development as neurons generated from the VZ migrate radially to the cortical plate using these processes as a guide (Rakic 1971, 1972; Hatten and Mason 1990). After the onset of neurogenesis, RGCs generate another specialized cell type, intermediate progenitor cells (IPCs) – also known as basal progenitor cells or non-surface dividing cells (Haubensak et al. 2004; Miyata et al. 2004; Noctor et al. 2004). IPCs divide symmetrically either in the basal VZ or in the SVZ to produce two neuronal daughters, thereby potentially doubling the numbers of generated neurons. Although they are the progeny of RGCs, IPCs lack both apical endfeet and ascending processes and also differ from radial glia in their gene expression (Gotz and Huttner 2005). Thus, prior to the onset of mammalian embryonic neurogenesis, the telencephalic neuroepithelium diversifies to include RGCs which can self-renew while generating IPCs and neuronal daughter cells. However, whether these changes fully represent the complexity of the VZ and are sufficient to generate the overall diversity of the maturing telencephalon is still a matter of debate.
From the earliest descriptions of the neocortical neuroepithelium to the present, there has been contention over its composition. Light and electron micrographs revealed mitotic cells at the ventricular surface that lacked cell processes and therefore appeared to be morphologically distinct from radial glia (Sauer 1935; Stensaas and Stensaas 1968; Hinds and Ruffett 1971). In contrast, tritiated ([3H]) thymidine injection appeared to uniformly label cells in the telencephalic proliferative zones (Sidman et al. 1959; Fujita 1963) suggesting a homogeneous precursor population. Subsequently, immuno-electron microscopy (EM) revealed the presence of glial fibrillary acidic protein (GFAP)+ cells intermixed with GFAP− cells in the VZ of rhesus monkeys (Levitt et al. 1981, 1983), pushing the argument back towards a heterogeneous population of precursors.
More recently, the identity and character of telencephalic VZ cells have been elucidated by comprehensive studies using increasingly sophisticated techniques. Many of these advances have been made possible by the discovery of several markers which are enriched in prenatal murine RGCs, including vimentin (Dahl et al. 1981), brain lipid binding protein (BLBP) (Feng et al. 1994), and astrocyte-specific glutamate transporter (GLAST) (Shibata et al. 1997). Several technological advances have enabled high-resolution examination of the viable embryonic brain. For example, time-lapse fluorescence imaging and genetic fate-mapping studies have conclusively demonstrated that RGCs can generate not only astrocytes and oligodendrocytes, as originally thought, but also neurons and neuronal precursors (Noctor et al. 2001, 2004; Malatesta et al. 2003; Anthony et al. 2004). Through these compelling results, RGCs are posited to be the sole residents of the VZ and have been heralded as the source of most cortical cells – at least in rodents.
Despite the comprehensive nature of these time-lapse and fate-mapping studies, several in vitro and in vivo analyses suggest that the mammalian VZ may include multiple different types of precursor cells, perhaps even multiple types of RGCs. For example, both in vivo and in vitro studies have shown that radial glia as a population can generate neurons, astrocytes and oligodendrocytes, but that very rarely will any individual RGC give rise to all three (Luskin et al. 1988; Grove et al. 1993; Qian et al. 1998, 2000; Malatesta et al. 2000b, 2003; McCarthy et al. 2001; Noctor et al. 2001, 2004). Furthermore, data generated in rodents demonstrates that expression of antigens and surface eptiopes – RC2, GLAST, BLBP, JONES, A2B5, cholera toxin, tetanus toxin, and combinations thereof – can discriminate among subsets of RGCs and neural progenitors which differ in abundance and lineage potential throughout neurogenesis (Hartfuss et al. 2001a; Maric et al. 2003, 2007).
In the embryonic human telencephalon, it has been established that RGCs are but a subgroup of the VZ population. Using immunohistochemistry, Zecevic (2004) detected a subpopulation of vimentin+ RGCs in humans that express the neuronal marker SMI-31 (phosphorylated neurofilament) at six gestational weeks (g.w.). At this same early developmental stage, other VZ cells, including mitotic cells at the ventricular surface, express the neuronal markers βIII-tubulin, MAP2, and SMI-31, but do not express conventional RGC markers. Howard et al. (2006) showed that dividing 4A4+ RGCs and SMI-31+ neuronal progenitors are simultaneously present in the telencephalic VZ at the onset of neurogenesis. However, SMI-31+ cells appear in the telencephalon before 4A4+ cells do (4.5 g.w. vs. 5.5–6 g.w. respectively). Altogether, these data depict a VZ composed of at least three classes of progenitor cells in humans: RGCs that express only glial markers, multipotent bipolar precursors that co-express glial and neuronal markers, and committed neuronal precursors that only express neuronal markers. This raises the question as to whether the degree of heterogeneity in the VZ is greater in higher vertebrates such as primates than in lower vertebrates like rodents, reptiles, fish, and birds. It seems the answer may be related to the tools used to ask the question.
Our recent studies (Gal et al. 2006) indicate that similar to the human telencephalon, the murine VZ also contains multiple precursor cell types. Employing a multidisciplinary approach, we have found two VZ cell types in mouse embryonic telencephalon which differ morphologically and by gene expression. Using in utero electroporation of membrane-tagged enhanced green fluorescent protein (EGFP) (pEGFP-f) to label cells dividing at the ventricular surface, RGCs that maintained a basal, ascending process were found alongside dividing cells lacking an ascending process. This latter cell type was named short neural precursors (SNPs). To conclusively demonstrate the presence of both RGCs and SNPs within the VZ, ultra-thin (70 nm) serial sections of the mouse and monkey VZ were imaged using EM, followed by 3D reconstruction of the EM micrographs. It is important to note that reconstructions of dividing VZ cells expressing the EGFP-f protein were performed to ensure that the thin and wispy membranes of dividing RGCs would not be missed. These reconstructions clearly identified mitotic cells lacking ascending processes. In addition, in utero electroporation with cell type-specific promoters showed that the GLAST and BLBP promoters can drive reporter gene expression in mitotic RGCs. In contrast, reporter gene expression from a tubulin α-1 (Tα1) promoter, which is also expressed by neuronal progenitors and post-mitotic neurons, occurs only in dividing SNPs. Thus, electroporation with these promoter constructs provided a molecular means of distinguishing between various VZ subtypes. Importantly, this study also revealed that Tα1-expressing SNPs remain proliferative within the VZ for at least a full cell cycle after electroporation, suggesting that SNPs differ from the intermediate progenitors (IPCs) which quickly exit the VZ and migrate to the SVZ (Noctor et al. 2004). Thus, sampling of the murine VZ cell population using in utero electroporation elucidates multiple types of precursor cells. The reason why SNPs have not been identified in studies using retroviral infection are unclear, but may be related to selective expression of the cytomegalovirus DNA promoter or to differential susceptibilities of VZ cell types to retroviral infection.
Further study of the multiple VZ cell types, including the RGC and SNP populations, is needed to more fully characterize the degree of heterogeneity in the mammalian VZ. Future efforts will take advantage of the differential promoter expression found between RGCs and SNPs to uncover potential physiological (e.g. membrane physiology, cell cycle parameters, etc.) and lineage potential differences between subtypes of VZ cells. Indeed, our recent studies (Mizutani et al. 2007; also see section on ‘Notch signaling in neural progenitor diversity’ below) used this differential promoter expression to shed new light on the signaling pathways that may control VZ cell lineage development. In particular, RGCs and SNPs appear to differ in their requirement for Notch pathway signaling. Furthermore, it has become clear that the RGC population is in itself diverse. Important differences have been found between RGC subtypes, including particular requirements for signaling factors that maintain the elongated RGC morphology (see section on ‘Function and signaling in radial glia’, below).
Notch signaling in neural progenitor diversity
The Notch signaling pathway plays a fundamental role in the regulation of neural progenitor cell development. Such a role was first suggested by the ‘neurogenic’ phenotype associated with loss of Notch signaling in fruit flies (drosophila); when the Notch cascade was disrupted, too many neurons were generated (Louvi and Artavanis-Tsakonas 2006). Further analysis of this phenotype revealed that Notch signaling regulates numerous binary fate choices during drosophila neural development (Fig. 4). For example, Notch is first used to select a cell to be a sensory organ precursor in the PNS (Guo et al. 1996; Gaiano and Fishell 2002). During subsequent rounds of division, Notch regulates specification of type IIa and IIb precursors from the sensory organ precursor, and finally the binary choices between pairs of cells that will become the socket and bristle, or the sheath and sensory neuron. This pattern of cell fate specification shows that Notch signaling can be used repeatedly in a neural lineage to make numerous cell fate choices, including the choice between two different types of proliferating neural cell types.
In mammals, the Notch signaling pathway includes four receptors (Notch1–4), and five classical ligands (Delta-like1,3,4 and Jagged1,2) (Yoon and Gaiano 2005). These are all single pass transmembrane proteins that permit signaling between adjacent cells through direct contact. Ligand activation results in the intramembranous cleavage of the Notch receptor by the γ-secretase complex, and the release of the intracellular domain, which then translocates to the nucleus and associates with the DNA-binding protein C promoter-binding factor 1 (CBF1) (Selkoe and Kopan 2003). In the nucleus, the NICD/CBF1 complex activates the expression of target genes such as the bHLH transcription factors encoded by the Hes and Hey genes (Iso et al. 2003). Those genes then function to suppress neurogenesis by antagonizing the function of proneural transcription factors including Mash1 and the Neurogenins. In addition, the existence of non-canonical, potentially CBF1 independent, Notch receptor function has been suggested by numerous studies, although the mechanism of such non-canonical Notch signaling is very poorly understood (Martinez Arias et al. 2002) (Fig. 4).
Our understanding of Notch pathway function in vertebrates began in the mid-1990s with the observations that Notch receptor activation could inhibit neurogenesis and myogenesis, both in cell lines and in Xenopus embryos (Kopan et al. 1994; Nye et al. 1994). With respect to the nervous system, that work laid the foundation for what has been the prevailing view regarding Notch function in vertebrates: Notch inhibits neurogenesis and maintains a neural progenitor state (Yoon and Gaiano 2005; Louvi and Artavanis-Tsakonas 2006). The initial examples of this included the inhibition of neuronal differentiation by P19 cells (Nye et al. 1994), and the inhibition of primary neurogenesis in Xenopus by over-expression of the Notch ligand Delta (Chitnis et al. 1995).
The ability of Notch activation to inhibit neurogenesis in vertebrates, considered together with the neurogenic phenotype of Notch pathway disruption in drosophila, suggested conservation of function. The role of Notch during vertebrate neural development was largely modeled not after the idea that Notch might regulate binary fate choices, but instead after the idea that its primary role was to maintain the proliferative population. Such a role is more akin to how Notch functions in the drosophila CNS, where neuroblasts are maintained, or ‘self-renewed,’ while generating successive rounds of more specialized daughter cells called ganglion mother cells, which divide again to generate neurons and/or glia (Gaiano and Fishell 2002). The nature of the vertebrate neural progenitor pool, where a large group of proliferating cells is maintained while a subset undergoes neuronal differentiation, was more readily compared with drosophila CNS development. Subsequently, the ideas that Notch might regulate binary fate choices in the vertebrate nervous system, as it does in the drosophila PNS, or that Notch might be used iteratively during the generation of progressively restricted progenitor cell types, as it does in both the drosophila PNS and CNS, received little attention.
In recent years our understanding of Notch function during neural development has only modestly changed. Some time ago it was determined that in addition to maintaining progenitor fate, Notch signaling could promote glial fate, and in particular radial glial and astroglial fate (Gaiano et al. 2000; Chambers et al. 2001; Gaiano and Fishell 2002). However, concurrent with these findings others showed that radial glia are embryonic neural progenitors (Malatesta et al. 2000a; Noctor et al. 2001), and that certain groups of cells with astrocytic character in the adult brain are neural stem cells (Doetsch et al. 1999; Laywell et al. 2000; Seri et al. 2001). Thus, the ability of Notch to promote certain glial cell types appeared to be at least partially related to its ability to promote progenitor character. Nevertheless, this work suggested that cells with Notch activation did not simply maintain the status quo, but could also undergo phenotypic changes as a result of pathway activation.
Numerous studies have demonstrated that Notch signaling plays a critical role in regulating the progenitor pool during neocortical development (Hitoshi et al. 2002; Yoon et al. 2004; Yoon and Gaiano 2005). First, Notch pathway components, including targets like Hes1 and Hes5, are expressed in many cells throughout the VZ (Yun et al. 2002; Mason et al. 2005). Second, gain-of-function studies have found that activating the pathway inhibits neuronal differentiation, and can promote the maintenance of progenitor character (Gaiano et al. 2000; Hitoshi et al. 2002; Yoon et al. 2004; Mizutani and Saito 2005). Loss-of-function studies, while often complicated by early lethality and pleiotropy, have supported this view (Yoon and Gaiano 2005). In general, a long-standing model has persisted regarding how Notch functions in mammalian neural development, and challenges associated with study of this process, have led to some stagnation in the field.
The traditional model of Notch function during mammalian corticogenesis, while supported by both gain- and loss-of-function experiments, is nevertheless almost certainly over-simplified. As discussed throughout this review, neocortical progenitors are not a homogenous pool of cells charged only with making the decision to become a neuron or not. Instead the neocortical germinal zone is comprised of numerous distinct proliferative cell types (Grove et al. 1993; Hartfuss et al. 2001b; Haubensak et al. 2004; Gal et al. 2006; Mizutani et al. 2007; Pontious et al. 2008). The most obvious distinction is between those progenitors located in the VZ and those located in the SVZ. The decision of some daughters of VZ cell divisions to migrate into the SVZ, where they will undergo subsequent divisions as basal progenitors, is a choice that is highly likely to be regulated. Furthermore, based upon gene expression and morphological analyses, it is clear that not all VZ cells are the same. For example, while Tbr2 is typically described as a marker of basal progenitor character, and is expressed throughout the SVZ, during mid-neurogenesis Tbr2 is also expressed in many VZ cells (Englund et al. 2005). Such VZ cells may be in the process of transitioning into SVZ cells, but as some were found to incorporate BrdU while in the VZ, these are at least temporarily VZ progenitors.
The existence of neocortical progenitor heterogeneity raises the questions: how is it generated, and what sorts of signaling differences exist between distinct progenitor subtypes? Prominent among the molecular cascades known to create heterogeneity in an initially homogeneous pool is the Notch pathway (Louvi and Artavanis-Tsakonas 2006). As the full complement of Notch pathway components, including ligands, receptors, modulators, and targets, is expressed in the telencephalic VZ (Lindsell et al. 1996; Zhong et al. 1997; Irvin et al. 2001; Yun et al. 2002), this pathway is a prime candidate to generate progenitor heterogeneity in that region. Several genetic studies, although focused on the LGE, have provided evidence that Notch signaling is not uniformly utilized by all telencephalic progenitors within a given region. One such study compared the phenotypes of numerous mouse mutants, including those with disrupted Dll1 or Mash1, and evaluated progenitor pool composition and the cell types generated from those progenitors (Yun et al. 2002). That work constructed a model indicating that there were three different progenitor subtypes, P1 and P2 in the VZ, and P3 in the SVZ. Furthermore, they suggested that both P1 and P2 could be inhibited from progressing to the next progenitor state by Notch activation. However, the authors noted that while P1 progenitors did not express Mash1, consistent with repression by the Notch-CBF1-Hes cascade, P2 progenitors were Mash1+. These findings raised the possibility that Notch inhibits the progression from P2 progenitor in the VZ to P3 progenitor in the SVZ without activating the Hes genes. More recent work has provided direct support for the idea that Notch may inhibit the differentiation of certain neocortical progenitors without operating through the traditional CBF1-Hes cascade (see below) (Mizutani et al. 2007).
A second study that disrupted Notch1 in the telencephalon observed a substantial reduction in the generation of early born striatal patch neurons, but not in later born matrix neurons (Mason et al. 2005). This finding was unexpected, as the traditional model of Notch function in neural development suggests that precocious progenitor differentiation at early time points would lead to a more significant reduction in later born cell types. One plausible interpretation of these findings is that there are two progenitor subsets early in striatal development: one that is Notch1 dependent and divides asymmetrically to generate patch neurons, and another that is Notch1 independent and divides symmetrically in preparation for generating matrix neurons later in development. How such progenitors relate to the P1, P2, and P3 progenitors described in the model above is not clear. However, both studies support the existence of progenitor heterogeneity with respect to Notch signaling.
More recently, several studies using transgenic reporter mouse lines have found that the canonical Notch-CBF1-Hes cascade is indeed not uniformly utilized throughout the telencephalic VZ (Ohtsuka et al. 2006; Basak and Taylor 2007; Mizutani et al. 2007). The notion that Notch pathway activation occurs in only a subset of neural progenitors at any given time is not a novel idea. Indeed, the traditional model regarding pathway function predicts that Notch ligand expressing cells will be present in a ‘salt and pepper’ pattern interspersed among cells that have higher levels of Notch receptor activation. Such heterogeneity has generally been attributed to the belief that the subset of cells in the VZ expressing high levels of ligand are differentiating neurons. Those cells would activate Notch receptors on neighboring VZ cells, thereby inhibiting their differentiation at that time. While this may well be the mechanism used to generate neurons from a certain subset of progenitors, during neural progenitor pool diversification, some ligand-expressing cells could be a newly specified progenitor subtype that is activating Notch on its neighbors to maintain or promote a distinct progenitor subtype.
Consistent with the existence of Notch signaling heterogeneity in the telencephalic progenitor pool, one study found that the Hes5 promoter (a target of Notch-CBF1 activity) is not uniformly active in the VZ (Basak and Taylor 2007). In this study, analysis of a transgenic mouse line with GFP driven by the Hes5 promoter showed that GFP expression was distinctly absent from a subset of VZ cells. While it is possible that some of those cells were newly generated neurons, the often clustered organization of the cells and the fact that some were found to express the progenitor markers Nestin, GLAST, and Pax6 supported the idea that they were progenitors. Thus, this study identified heterogeneity among VZ progenitors with respect to Hes5 promoter activity. Interestingly, however, that work also found that only GFP+ cells could form neurosphere colonies in vitro, suggesting that progenitors without Hes promoter activity had limited proliferative activity.
Our recent studies, also using a transgenic reporter strategy, provided additional support for heterogeneity in Notch utilization in the telencephalic VZ, and also suggested that the modulation of the Notch cascade was causally connected to generating that heterogeneity (Mizutani et al. 2007). This work used a reporter transgene [called transgenic Notch reporter (TNR)] designed to drive GFP expression from a responsive element including CBF1 binding sites derived from Epstein-Barr virus. The pattern of transgene activation in the embryonic telencephalon in this work was quite similar to that described above using the Hes5 promoter (compare Fig. 1c in Mizutani et al. 2007; to Fig. 3a in Basak and Taylor 2007), further supporting the notion that Notch-CBF1 targets are not uniformly utilized in the telencephalic VZ. One noteworthy difference between our TNR study and that of Basak and Taylor, is that while the former was able to generate neurosphere from GFP− cells, the latter was not. The reason for this difference remains to be determined.
Beyond evaluating the status of Notch-CBF1 activity in telencephalic progenitors, our TNR study also examined the relationship between that activity and gene expression driven from the radial glial GLAST and neuronal lineage Tα1 promoters [indicative of RGC and SNP character, respectively, as discussed above (Gal et al. 2006)]. In this study, it was found that GFP+ cells primarily drove expression from the GLAST promoter, while GFP− cells primarily drove expression from the Tα1 promoter. These results suggested that VZ progenitors with reduced Notch-CBF1 activity were more neurogenic, a notion supported by both in vitro and in vivo differentiation assays (Mizutani et al. 2007).
Our TNR study went on to show that the reduced Notch-CBF1 activity in SNPs [called ‘intermediate neural progenitors (INPs)’ in that work], resulted from an active and heritable block to signal transduction. While the mechanism of that block remains to be elucidated, the evidence was provided that it was causally linked to the transition from RGC to SNP/INP: shRNA-mediated knockdown of CBF1 shifted VZ progenitors from the former to the latter. This work also suggested that Notch receptor activation could inhibit the differentiation of both RGCs and SNP/INPs, although in the case of the latter with attenuated CBF1 signaling. As discussed above, the existence of two types of VZ cells maintained by Notch, but potentially using different mechanisms, was predicted by a model devised from a previous genetic study (Yun et al. 2002).
In summary, the work discussed in this section strongly supports the idea that distinct subsets of neural progenitors utilize the Notch signaling pathway differently. In addition, there is evidence that regulation of the Notch signaling cascade may play a role in the progression from one proliferative neural cell type to the next, at least in the context of the neocortical VZ. As our understanding of neural progenitor diversity expands it will be interesting to determine to what extent other signaling pathways known to regulate progenitor behavior (e.g. Wnts, fibroblast growth factors, and BMPs) are differentially utilized and/or play a role in the generation of progenitor diversity. The interplay of Notch signaling and the generation of cell diversity via combinatorial transcription factor expression also remains an important and relatively unexplored question.
Function and signaling in radial glia
During cortical development, radial glia serve as the primary scaffold for migrating cells that form the cortical plate. In addition, as described above, recent work has excitingly revealed that these elongated cells are also neural progenitors, producing neurons and glia that will populate cerebral cortex. As the cortical plate develops and migrating neurons are produced, specific signals maintain radial glia in their proper elongated morphology. Although our understanding of the complete set of these signals is still evolving, several substances originate from migrating neurons, which act on various structural molecules within the radial glia, maintaining their extended shape. Other cells in the marginal zone of the developing cortical plate secrete substances that are important to the positioning of migrating neurons and the morphology of radial glia.
Both genetic mutations and environmental insults can influence development of the cerebral cortex via affecting the proper morphology of RGCs and subsequent neuronal migration. In one type of cortical dysplasia induced by an environmental toxin, both neuronal migration and the morphology of radial glia are disrupted. In this model, an antimitotic agent [methylazoxy methanol (MAM)] is administered to pregnant ferrets early during corticogenesis (E24), which results in the elimination of a population of cells destined to become one of the first generated cortical layers in the ferret. Earlier work in this field indicates that this treatment results in animals with disorganized radial glia and poorly formed cerebral cortex (Noctor et al. 1999). It is unlikely that the radial glial disruption is a result of toxic poisoning, as administering MAM at later dates during corticogenesis, but at times of active migration and elongated radial glial morphology, does not result in any changes in this cell type (Noctor et al. 1999). The morphology of the disrupted radial glia can be rescued by coculture with normal cortex, suggesting that this structure contains a radializing factor competent to restore elongated morphology; i.e. this results in an undifferentiated state capable of continued proliferation and support of migration. Coculture with normal cortical plate also improves migration of neurons into their proper position (Hasling et al. 2003).
The treatment of E24 MAM treated cortex with normal cortical plate results in two basic outcomes: maintaining the elongated morphology of radial glia and the migration of neurons into the cortical plate. It is likely that the processes acting on radial glia to maintain them in an undifferentiated state include signals arising from neurons, leading to cascades that produce structural alterations of the radial glial phenotype. Migrating neurons and radial glia are further influenced by factors existing in, or en route to, the cortical plate. Failure of RGCs to maintain their proper morphology and proliferative state can lead to numerous disorders in which neurons fail to migrate properly and result in varying degrees of cortical dysplasia (Poluch and Juliano 2007).
One of the factors produced by migrating neurons that influence radial glial morphology is neuregulin (NRG1) (Anton et al. 1997; Rio et al. 1997; Gierdalski et al. 2005). NRG1 signals through erbB receptors and affects the morphology of radial glia through a number of mechanisms. In the model of MAM treatment, the addition of exogenous NRG1 to organotypic cortical cultures of E24 MAM cortex causes dramatic reorganization of disorganized radial glia to a more elongated appearance. NRG1 is normally present in developing ferret cortex, but diminished in the MAM treated model. To further understand the mechanism of NRG1 action, we (Gierdalski et al. 2005) tested the effects of activating or blocking erbB tyrosine kinase receptors, which are involved with NRG1 processing in other models (Fig. 5). In developing ferret cortex, NRG1 acts through erbB receptors, confirmed by evidence that antibodies stimulating erbB receptors result in improved radial glial morphology. In addition, using a soluble form of erbB receptor subtype that binds to available NRG1, does not allow improvement of radial glial morphology (Gierdalski et al. 2005).
These observations suggest a means of activation on radial glia through migrating neurons that express NRG1, which in turn act on the erbB receptors present on radial glia. The complete signaling pathway by which NRG1 and erbB receptors cooperate to influence the elongation of radial glia is complex and not completely known. ErbB receptors, however, are also activated by signaling through the Notch pathway, which further contributes to maintaining radial glial morphology (Fig. 5). Notch is typically described to be activated by the ligands delta and jagged that originate in migrating neurons. Notch activation can regulate radial glial phenotype both through erbB production and up-regulation of BLBP, a hydrophobic ligand-binding protein present in both the nucleus and cytoplasm (Gaiano et al. 2000; Gaiano and Fishell 2002; Patten et al. 2003, 2006; Ever and Gaiano 2005).
The signaling cascades involving Notch, erbB receptors, and NRG1 are important for mediating the elongated phenotype of radial glia, and presumably maintain these cells in a proliferative state. These pathways may or may not influence the process of neuronal migration, although impaired radial glial morphology almost certainly impedes this ability. Two of the main disruptions after E24 MAM treatment include disorganized radial glia and failure of neurons to migrate appropriately into the cortical plate. The exogenous application of NRG1 restores radial glial morphology, but despite a dramatically improved radial glial phenotype after NRG1 treatment, newly born neurons cannot migrate effectively into the neocortex. In contrast, earlier evidence revealed that coculture with normal cortex leads to improved migration, indicating that an additional factor in the cortical plate is important for proper migration (Hasling et al. 2003). Thus, an elongated radial glial phenotype alone is not sufficient restore migration, and depends on other factors.
As such, the effects of reelin have also been explored. Reelin is important in mediating radial glial phenotype and in cortical migration, although its precise role in both these functions continues to evolve (Soriano and Del Rio 2005; Forster et al. 2006; Cooper 2008). For example, MAM damaged cortex treated with a reelin source improves radial glial morphology, as well as migration into the cortical plate (Schaefer et al. 2008). In addition, this same study revealed that radial glia can be elongated in the absence of reelin, although neurons do not migrate properly without reelin present. In contrast, wild type cortical explants restore both radial glial morphology and migration suggesting that normal cortical plate contains the elements crucial for maintaining radial glial morphology and for proper migration (Schaefer et al. 2008). Reelin acts through an established signaling cascade involving apoliprotein E receptor-2 (ApoER2), very low density lipoprotein (VLDL) receptors and Dab1 (Jossin et al. 2003). If elements of this signaling cascade are interrupted, the reelin meditated improvement in radial glial morphology or in migration of cells into the neocortex does not occur in E24 MAM treated ferrets. These findings strongly implicate reelin as an important factor involved in producing both elongated radial glial morphology and migration of neurons into the cortical plate.
In addition to the recent evolution of our understanding of radial glia as neural progenitors, several studies demonstrate that these cells are not homogeneous but have diverse properties (Pinto and Gotz 2007). Distinguishing characteristics can be revealed in part by differential expression of specific radial glial markers; identifying features among different populations of progenitor cells are beginning to emerge. In the mouse, RGCs produce different combinations of GLAST, RC2, and BLBP; distinct functions may exist for each combination (Pinto and Gotz 2007). It is also possible that unique patterns of expression are a function of developmental switches and expressed differentially as the cortex matures (Anthony et al. 2004; Ever and Gaiano 2005). In the ferret, radial glial markers are not identical to those present in the mouse. Vimentin is a strong radial glial marker in this species (Voigt 1989; Noctor et al. 1999), the classic marker RC2, however, is specific to the mouse and not expressed in ferrets. Several other markers present in the mouse (such as GLAST and nestin) do not appear to be strongly represented in ferret RGCs, but another well-known marker, BLBP, is strongly immunoreactive in ferret radial glia. In addition, this indicator appears to identify a separate and distinct population from those immunoreactive for vimentin.
Our further analysis of the BLBP immunoreactive population of radial glia revealed surprising features. Analysis of this set of cells indicated that although they were slightly disrupted after MAM treatment, the amount of radial glial disorganization was minimal, especially compared with that observed in cells labeled with vimentin (Poluch and Juliano, unpublished observations). To further assess if different factors influenced vimentin-expressing versus BLBP-expressing radial glia, essential components of two pathways known from previous studies to be important in radial glial morphology in the ferret (the reelin and NRG1 signaling pathways) were evaluated.
As stated above, blocking elements of the reelin signaling pathway including the VLDL/ApoER2 receptors and Dab1 prevents the repair of MAM treated radial glia in the presence of reelin. Interestingly, this function is only for the vimentin immunoreactive radial glia. The BLBP expressing radial glia, in contrast, were not affected by interrupting the reelin signaling pathways, they remained elongated, with morphology only slightly disrupted in the MAM treated brain (Poluch and Juliano, unpublished observations). This suggests that the RGCs expressing vimentin are responsive to reelin, while the BLBP expressing radial glia are not, as they are not altered by blocking elements of reelin signaling. Phosphoinositide 3-kinase (PI3K) is an element common to both reelin and NRG signaling (Fig. 5) (Beffert et al. 2002; Citri et al. 2003; Jossin and Goffinet 2007; Kanakry et al. 2007). Blocking PI3K in the presence of a normotopic source of reelin (using a stably transfected line of cells expressing reelin), demonstrated that the vimentin expressing radial glia are disorganized, i.e. not repaired by the presence of reelin, and the BLBP expressing radial glia are also disrupted. If the same experiment is conducted in the presence of NRG1, the vimentin expressing radial glia remain disrupted, while the BLBP immunoreactive cells are elongated and exhibit the slightly affected morphology seen after MAM treatment. This suggests that the exogenous NRG1 is not able to rescue the specific population of vimentin expressing cells from the PI3K blockade, while the BLBP expressing cells remain relatively elongated.
Consistent with the notion as described above that the VZ is highly heterogeneous, these findings suggest the existence of at least two distinct populations of RGCs that respond differently to signals. One set of radial glia (the vimentin expressing) are impaired by PI3K blockade so that they cannot be repaired by the presence of either reelin or NRG1. The other set of radial glia (the BLBP expressing) are impaired by PI3K blockade in the presence of reelin, but are not affected in the presence of NRG1, suggesting that they make use of parallel pathways involving erbB receptors and interactions with Notch, leading to increased BLBP production (Fig. 5). BLBP may be regulated by migrating neurons and expressed while radial glia are used as a scaffold for migration (Feng et al. 1994; Feng and Heintz 1995). Reelin also appears essential for proper migration and the two substances may work in tandem where BLBP expression is crucial to the initiation of migration as neurons leave the VZ, while reelin mediates the final stages of cortical laminar fate.
As reviewed above, progenitor cells in the mammalian telencephalon are highly diverse. This diversity is generated by a combination of progenitor pool specific transcription factor expression and Notch signaling cascades. Although many of the mechanisms of exactly how diverse subtypes of neurons and glia are generated from uncommitted progenitors remain unknown, much has been learned over the recent years. Most importantly, unraveling the mechanisms of neural progenitor cell diversity in the brain has tremendous clinical importance. First, defects in any of these processes can have devastating and long-lasting consequences on brain function. For example, although varied in their phenotypes, pathogenesis and etiology, work in both animals and humans indicates that the abnormal development of interneurons may be an underlying causative factor, or contribute to the phenotype of a variety of developmental disorders (Rubenstein and Merzenich 2003; Levitt et al. 2004). These include complex and widespread disorders such as autism spectrum disorders, schizophrenia, Fragile X Syndrome and Down syndrome, as well as more rare single gene disorders such as a subtype of X-linked lissencephaly known as X-linked lissencephaly with abnormal genitalia. Second, as we begin to understand how diversity is generated during embryogenesis, this will pave the way for generating neurons and glial cells for cell replacement strategies in both disease and injury states for restoration of normal neuronal/glial function. Ideally, specific neural subtypes could be generated from endogenous resident neural stem cells in the adult brain or embryonic stem (ES) cells, and then either coaxed to differentiate into the appropriate subtypes in situ or first differentiated in vitro and surgically transplanted directly into the injured/diseased brain. However, while such an approach provides a promising area of pursuit, a number of caveats that must be taken into consideration. First, although functional integration of new neurons is an ongoing process that occurs in both the post-natal olfactory bulb and hippocampus, it remains unknown if adult progenitor cells can be utilized to generate other functional mature neural cell types. In contrast, transplantation studies of embryonic neural progenitor cells in animals have generated some encouraging results with regard to the ability of exogenously delivered cells to become functionally integrated into the post-natal brain. For example, MGE progenitor cells isolated from the E12.5/E13.5 telencephalon can differentiate into interneurons and integrate into early post-natal cortical circuitry in an apparently normal manner (Alvarez-Dolado et al. 2006). Thus, as prenatal progenitor cells already have an endogenous capacity to generate interneurons, it is also possible at least in theory, to utilize these cells for replacement strategies.
Much attention has also been placed on the utility of ES cells as a source of cells for regeneration both within and outside of the nervous system (Taylor and Minger 2005). Although holding much promise, up until this point differentiation of ES cells into functional neurons in vivo has proven to be difficult. Studies of a number of cell types, including retinal rods (MacLaren et al. 2006), spinal cord motoneurons (Wichterle et al. 2002), and midbrain dopaminergic cells (Roy et al. 2006) reveal that new cells can only incorporate into host tissue only if they are differentiated (or allowed to differentiate) along their normal developmental pathway. Thus, understanding the normal developmental program of individual neural subtypes for experimental approaches as described above provides an essential framework for the design of rational ES-derived cell replacement studies.
Thus, the issue of whether there are molecular, morphological, and physiological differences between embryonic VZ and the pathways required for their specification and differentiation is essential for the rational design of effective therapeutic strategies for congenital malformations and syndromes as well as repair of the injured or diseased brain.