PRONEURAL GENES ARE NECESSARY AND SUFFICIENT TO INDUCE NEUROGENESIS
Creating a functional nervous system is a daunting task, requiring that hundreds of neuronal subtypes are born at precise times and in specific locations. Remarkably, one class of proteins, the proneural factors, is used over and over again, not only to specify cells to become neuroblasts but also to promote the differentiation of all neuronal subtypes.
The proneural factors are a family of tissue-specific transcription factors that share a conserved basic helix–loop–helix (bHLH) motif, where the basic domain contacts DNA, and the HLH domain binds to ubiquitous bHLH partners, the E proteins. Proneural protein-E protein heterodimers bind regulatory DNA at E-box sites (CANNTG) and activate transcription of neuronal genes. The proneural family is sufficient and necessary for neuronal development. Ectopic expression of some family members is sufficient to convert naive neuroectoderm into differentiated neurons (Lee et al., 1995; Ma et al., 1996; Takebayashi et al., 1997). Moreover, Drosophila and mice lacking one proneural gene, or two redundant proneural genes, fail to develop subsets of neuronal precursors and differentiated neurons (Jarman et al., 1993; Ma et al., 1998; Fode et al., 2000; Inoue et al., 2002). Although it is clear that proneural genes are critical regulators of neuronal determination and differentiation, their multifaceted roles in this complex process are still being discovered.
PRONEURAL GENES IN DROSOPHILA NEUROGENESIS
Proneural factor function is particularly well understood in the context of sensory organ precursor (SOP) development in the Drosophila peripheral nervous system (PNS). The bHLH factors achaete (ac) and scute (sc) from the Achaete-Scute Complex (AS-C) (ac, sc, lethal of scute, asense), atonal, and amos are each required for development of SOP subtypes. AS-C specifies external sense organs (bristles; Campuzano and Modolell, 1992); ato specifies chordotonal receptors, R8 photoreceptors, and a subset of olfactory sensilla (Jarman et al., 1993, 1994; Gupta and Rodrigues, 1997); and amos specifies two classes of ato-independent olfactory sensilla (zur Lage et al., 2003). SOP development initiates when small groups of cells, called proneural clusters (PNCs), express a proneural gene at low levels (Fig. 1A). PNCs are competent to become neuroblasts, but only when a single PNC cell increases proneural gene expression does a cell transition to becoming a neuroblast and delaminate from the PNC.
Selection of a neuroblast from the rest of the PNC is mediated by lateral inhibition. In the future neuroblast, proneural factors autoregulate their own expression as well as positively regulate expression of the Notch receptor ligand, Delta (Heitzler et al., 1996). Increased Delta activates Notch in adjacent cells, triggering an intracellular signaling cascade that leads to inhibition of proneural gene expression (Fig. 1B,C; Artavanis-Tsakonas et al., 1990). As a result, cells neighboring the future neuroblast ultimately acquire non-neuronal fates (Artavanis-Tsakonas et al., 1999). Drosophila neurogenesis is a complex process where tight control of both Notch signaling and proneural gene expression are required for specification as a neuroblast.
PRONEURAL GENES IN VERTEBRATE NEUROGENESIS
Events in vertebrate neural development parallel those in Drosophila. The vertebrate proneural genes fall into two classes: the Ash family are homologous to achaete-scute, and the Ath, Neurogenin (Ngn), and NeuroD subfamilies are homologous to atonal. Unlike in Drosophila, vertebrate proneural genes are often expressed in two waves. During primary neurogenesis in Xenopus, X-Ngn-R1 (Xenopus Ngn related) is expressed in broad stripes down along the neural plate. X-Ngn-R1 directly activates NeuroD, and NeuroD activates expression of neuronal differentiation markers in narrower stripes (Lee et al., 1995; Ma et al., 1996). The number of X-Ngn-R1–positive cells that differentiate as neurons is restricted by lateral inhibition. X-Delta-1 is expressed in neuroblasts, and inhibition of X-Delta-1 results in an expansion in the number of neural plate cells that differentiate as neurons (Chitnis et al., 1995). This finding shows that the role of X-Delta-1 in vertebrate neuroblasts is to prevent neighboring cells from adopting the neural fate, demonstrating that vertebrates and Drosophila use common mechanisms to initiate neurogenesis.
After a cell becomes a neuroblast, how do proneural factors regulate neurogenesis? Proneural factors lie at the interface of extrinsic (extracellular) signals and intrinsic (cell autonomous) signals that drive neurogenesis (Edlund and Jessell, 1999). Extrinsic signals render a cell competent to adopt the neuronal fate. For example, specification of motor neurons is initially dependent on exposure to sonic hedgehog (Shh; Ericson et al., 1996), and in cultured neural crest stem cells, the extracellular signaling molecule bone morphogenetic protein 2 (BMP2) induces expression of Mash1 (mouse ash1; Lo et al., 1997). Over time cells become independent of extracellular signals, maintain expression of proneural factors, and commit to the neural fate. This transition is most likely triggered by a cascade of intracellular events catalyzed by proneural factors, including autoregulation of proneural genes and expression of neuronal-specific genes (Mattar et al., 2004; zur Lage et al., 2004; Reeves and Posakony, 2005). The primer will first describe the intracellular events triggered by proneural factors. Next, the primer explores how extrinsic influences such as regional determinants, and also sequence differences between proneural factors, each contribute to specification of a vast array of neuronal subtypes.
PRONEURAL FACTORS ACTIVATE PAN-NEURONAL AND CELL TYPE-SPECIFIC TARGETS
A shared function among proneural factors is to endow a cell with general neuronal characteristics. That ectopic expression of many proneural factors can convert non-neuronal cells to neurons is an example of this early, shared role. There is emerging evidence that a “core” set of target genes may be activated by proneural factors (Fig. 2A; Brunet and Ghysen, 1999). Recently, two microarray studies were conducted to identify genes activated downstream of proneural factors. Genes activated by achaete-scute in Drosophila PNCs and SOPs are particularly enriched for metabolic enzymes (i.e., malate dehydrogenase, cytochrome P450), cytoskeleton remodeling factors (microtubule associated protein, ARF monomeric GTPase), and transcription factors (atonal, cato; Reeves and Posakony, 2005). Similarly, a screen for genes downstream of Ngn2 in mouse cortical neurons identified signal transduction components (Mef2c, EphA5), and cytoskeleton remodeling factors (protocadherin) (Mattar et al., 2004). These findings support the hypothesis that, during neurogenesis, core genes are activated that modify cell metabolism, rearrange cytoskeletal architecture, and activate signaling cascades. Expression of a common subset of genes changes the basic nature of the progenitor cell, and advances the adoption of basic neuronal characteristics.
In addition to initiating the neurogenic gene program, proneural factors promote differentiation to neuronal subtypes. Interestingly, there is apparently conflicting evidence that proneural factors function redundantly, and yet also have unique roles in specifying neuronal subtype identity. Functional redundancy has been demonstrated on multiple occasions; just two are listed here. Zebrafish ash1a and ngn1 each independently regulate development of pineal organ neurons (Cau and Wilson, 2003). In addition, Mash1 and Math3 redundantly specify neuronal fates both in mouse hindbrain neurons and retinal bipolar cells (Tomita et al., 2000). However, proneural factors cannot always functionally substitute for another, suggesting that they can acquire unique functions as well. In mice where Ngn2 coding sequence is replaced by that of Mash1, dorsal telencephalic neurons acquire ventral characteristics and express markers for GABAergic neurons, Dlx1 and GAD67 (although development of many neurons with modified gene expression is normal, again demonstrating some functional redundancy; Fode et al., 2000). In the same mice, Mash1 respecifies a subset of spinal motor neurons in the ventral spinal cord to interneurons expressing the marker Chx10 (Parras et al., 2002). Similarly in Drosophila, misexpression of ato can respecify SOPs to chordotonal organs (Jarman and Ahmed, 1998). Therefore, at least in certain contexts, proneural genes possess unique characteristics that result in activation of distinct target genes and development of particular neuronal cell fates. Mechanisms by which proneural factors acquire unique functional activities are discussed below.
MECHANISMS THAT GENERATE NEURONAL DIVERSITY
An enduring conundrum in the field of neurogenesis is how a limited number of proneural factors generate hundreds of neuronal subtypes. The two most basic models explaining this phenomenon are that proneural factor activity is dependent on (1) regional determinants (Fig. 2B) and/or (2) sequence-specific differences among proneural factors (Fig. 2C). Evidence supporting each of the models will be discussed below.
Proneural Functional Diversity Is Context-Dependent
Context-dependent proneural function is evident from the observation that the same transcription factor can promote development of diverse neuronal subtypes. Xath5 expression is limited to the pineal gland, retina, and olfactory placode and promotes both retinal ganglion and olfactory receptor cell fates (Kanekar et al., 1997; Burns and Vetter, 2002). Furthermore, forced expression of Xath5 at early time points in the retina promotes an early retinal fate, retinal ganglion cells, whereas forced expression at later time points promotes later fates, bipolar cells and photoreceptors (Moore et al., 2002). Proneural factors also directly activate neuronal subtype-specific genes. Math1 directly activates expression of the homeodomain gene Mbh1 in commissural neurons (Saba et al., 2005). Ath5 directly activates nicotinic acetylcholine receptor β-3 in chick retinal ganglion cells (Matter-Sadzinski et al., 2001). Taken together, the implication is that regionally expressed molecular determinants enable proneural factors to activate distinct subsets of genes.
Emerging evidence suggests that the outcome of proneural factor activation is dependent on a combination of regional determinants to confer patterning information to a cell (Fig. 2B). In the Drosophila central nervous system, subdivision of neuroectoderm is accomplished through epidermal growth factor receptor signaling in conjunction with the homeobox genes, ventral nervous system defective (vnd), an ortholog of vertebrate Nkx2.2, and intermediate nervous system defective (ind), an ortholog of Gsh1/2, to specify ventral and intermediate neuroblast subtypes, respectively (Buescher et al., 2002; Overton et al., 2002; Zhao and Skeath, 2002). Elements of this pathway are conserved in vertebrates. The ability of a population of ventral cells in the neural tube to differentiate as V3 intermediate neurons is dependent both on exposure to high levels of SHH, and on expression of Nkx2.2 (Briscoe et al., 1999). By contrast, formation of cells that lie more dorsally in the neural tube to become motor neurons and V2 interneurons is dependent on exposure to a lower level of SHH and expression of Nkx 6.1 (Roelink et al., 1995; Sander et al., 2000). These findings gave rise to the hypothesis that a homeodomain code specifies progenitor cell identity (Briscoe et al., 2000). How positional information results in activation of differing subsets of genes by proneural factors currently is unknown.
Proneural Functional Diversity Is Dependent Upon Sequence-Specific Differences
At least some aspects of functional diversity can be accounted for by differences in sequence between proneural factors (Fig. 2C). Two examples illustrating this point will be described here. First, specific protein domains in Mash1 and Math1 confer cell-type specific activities. Forced expression of Mash1 and Math1 in the chick neural tube each preferentially promotes formation of different neuronal cell types, dI3 and dI1 interneurons, respectively. It was revealed by chimeric protein analysis that helix 1 and 2 within the HLH motif of Mash1, and helix 2 within the HLH motif of Math 2, are required for these functional differences (Nakada et al., 2004). Second, sequence differences within the basic domains of Ato and Sc drive functional differences. The basic domains of Ato and Sc differ by 7 amino acids, and these nonconserved residues are particularly important for activation of chordotonal-specific gene expression by Ato (Chien et al., 1996). Of interest, in each of the two examples listed above, none of the residues that confer functional specificity are predicted to bind DNA, suggesting that they may instead associate with specific cofactors.
Neuronal Diversity May Be Mediated by Proneural Cofactors
An appealing model for generating neuronal diversity is that proneural factors bind cell type specific cofactors, and the combined activity of the two proteins induces neuronal subtype-specific gene expression (Fig. 2D). Evidence supporting this model comes from the finding that Ato and Sc bind different consensus E-box sites, where sequence specificity lies in sequence flanking the E-box. Moreover, multimers of Ato or Sc consensus binding sites drive reporter expression in Ato-dependent chordotonal neurons and Sc-dependent sensory neurons, respectively (Powell et al., 2004). These data suggest that Ato or Sc may drive neuronal subtype-specific gene expression by activating subsets of genes that have either Ato or Sc consensus E-box elements. Because the residues that contact DNA are conserved between Ato and Sc, the implication is that interactions with protein cofactors may promote binding to specific DNA sites.
Association with distinct cofactors could be dependent upon inherent differences between proneural factors, position-dependent environmental factors, or a combination of the two. Whereas generation of diversity by cofactors is an appealing hypothesis, few have been identified. Below are examples where there is a functional but no physical interaction, and a physical but no functional interaction, between proneural factors and interacting proteins.
There are two examples where bHLH factors cooperate with homeodomain proteins to promote specific neuronal fates. Math3 and Ngn2 each cooperate with LIM homeodomain proteins to promote development of motor neurons. Forced expression of Math3 or Ngn2 with both the LIM homeodomain proteins Lhx1 and Isl1 greatly enhances the number of motor neurons formed in the chick neural tube compared with overexpression of either bHLH or both LIM homeodomain proteins alone. Furthermore, chromatin immunoprecipitation experiments show that Math3, Lhx1, and Isl1 all bind directly to the promoter of a gene expressed in motor neurons, Hb9 (Lee and Pfaff, 2003). In a second study, it was shown that Mash1 and Math3 can each cooperate with the homeodomain protein Chx10 to make retinal bipolar cells (Hatakeyama et al., 2001). Both of these examples show that proneural factors can function together with specific homeodomain proteins to specify neuronal cell types. However, a physical interaction between the homeodomain and proneural proteins has not been shown. The mechanisms by which the two types of proteins work together to coordinate neuronal differentiation is not understood.
There is evidence of proneural factors interacting physically with other proteins, but a functional role for these interactions in generating specific neuronal cell types has yet to be determined. X-Ngn-R1 recruits the transcriptional activators p300 and PCAF (Koyano-Nakagawa et al., 1999). However, expression of the interacting proteins are not cell-type specific, and so are unlikely to confer neuronal subtype gene expression. In a yeast two-hybrid screen, Xenopus Xash1, Xath3, Xath5, and NeuroD each bind the homeodomain protein Six3 (Tessmar et al., 2002). Of interest, Six3 is expressed in a subset of neurons, but an in vivo functional interaction between proneural bHLH-Six3 proteins has not been examined. A rigorous screen for proneural cofactors has not been reported and may be a missing link to understanding how neuronal diversity is created.
Neuronal diversity is most likely generated by mechanisms dependent both on regional determinants and sequence differences. One possible model integrating both contributions is that structural differences between Atonal-related or Achaete-Scute–related subfamilies help confer functional diversity, but members within a subfamily are particularly dependent upon regional cues to generate differences. How structural differences and patterning information translates into neuronal diversity will undoubtedly be the subject of future research.
A CONVERSATION WITH THE EXPERTS
Current topics in the field of proneural factors in neurogenesis are discussed below in an interview with Andrew Jarman, Ph.D., Wellcome Senior Research Fellow and Reader in Developmental Cell Biology, Centre for Integrative Physiology, University of Edinburgh, United Kingdom, and Jane Johnson, Ph.D., Associate Professor, Center for Basic Neuroscience, University of Texas Southwestern Medical Center, Dallas (Fig. 3). An extended version of this discussion can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat.
Developmental Dynamics: How do you envision that proneural factors generate neuronal diversity? What is the relative contribution of sequence-specific differences compared to context-dependent differences?
Andrew Jarman: Clearly there must ultimately be differences in proneural factor function—the different proteins have similar but different intrinsic properties. atonal and scute drive the formation of different types of sense organ precursors. But these properties will be manifest in the context of patterning and signals. For example, atonal drives formation of different types of precursors in different locations. Intrinsic and extrinsic factors are often discussed as though they are exclusive alternatives. In reality of course, there is a complex interplay between the two.
Jane Johnson: I completely agree. Although signaling factors result in patterned expression of distinct bHLH factors, and presumably their cofactors, each bHLH appears to have a unique function in specification. When homologs of atonal and scute, Math1 and Mash1, are misexpressed in the chick neural tube, each causes an increase in a different cell type. However, there are certainly context-dependent differences in that, when overexpressed in other neural domains such as inner ear for Math1, you get different cell types than that found in the spinal neural tube, analogous to what Andrew stated happens with atonal.
Dev. Dyn.: What is the mechanism by which proneural factors generate diversity?
J.J.: Multiple groups have shown the importance of specific residues in the basic region and/or in the HLH domain of bHLH factors for their distinct functions. In each case, the important amino acids are those pointing away from the DNA or the E-protein heterodimer complex. The obvious interpretation is that there are factors interacting with the bHLH factors that provide the specificity. This could be in stabilizing the protein–DNA interaction or bringing in different transcriptional machinery.
A.J.: As Jane says, we know that the sequence of the bHLH domain is important in determining specificity (e.g., from Bassem Hassan's work and Jane's own work), but we don't know how. Differential protein interactions are the obvious interpretation, but the problem is that we don't have a clear example in which a protein cofactor actually interacts with the important amino acid residues in the bHLH domain to promote specificity. It hasn't been ruled out that the important amino acid differences affect DNA interaction rather more directly. We know that Scute and Atonal seem to use different E-box sequences to regulate at least some of their targets (our work). Do differential cofactor interactions govern the use of these different sites (as location of the “specificity residues” away from the DNA suggests), or vice versa?
Dev. Dyn.: Given the potential critical role for cofactors, is it surprising that so few have been identified?
J.J.: It is surprising that so few cofactors have been identified. However, I can think of all sorts of reasons why this is the case. It is possible that the cofactor recognizes the proneural factor/E-protein heterodimer, so approaches such as a yeast two-hybrid using only one half of the dimer would not work. It is possible that DNA binding is part of the complex or stabilizing the complex so immunoprecipitation might not yield cofactors.
Also rather than “cofactor,” the combination of sites on any given enhancer/ promoter, the developmental history (e.g., chromatin structure), etc., play into this. Another mechanism that probably overlays all of this is how each factor induces Notch signaling. This becomes even more complicated.
A. J.: My feeling is that there is no “magic bullet” universal cofactor for any particular proneural factor (otherwise we would have found them in genetic screens in Drosophila already). Instead there will be a complex set of interactions with different proteins that are important for different target genes in different contexts. For instance, we showed that Pointed protein interacts with Atonal to promote Atonal autoregulation. This interaction seems to be crucial for this particular target enhancer, but I don't think it will be required for Atonal's function on most of its other targets.
Dev. Dyn.: What do you think are some important questions in the field that remain to be answered?
J.J.: (1) Are there interacting factors and what are they? What complexes are forming and what are the specifics of these interactions? (2) What is regulating the activity of the bHLH factors: posttranscriptional and posttranslational. We know that the 3′untranslated region of Mash1 is required to obtain the endogenous expression pattern but we don't know how this is working. Phosphorylation and other modifications are likely important in regulating activity; what are these modifications and how are they being regulated? (3) We need a more comprehensive view of downstream targets for multiple bHLH factors in multiple tissues at multiple time. Bioinformatics would be useful to help sort and model the data.
A.J.: What is conserved across lineages in proneural gene function and what is lineage-specific? What do neurogenin-like genes do in Drosophila? What is the mechanistic basis of specificity? We're long on ideas but short on concrete detail.